14957, Use median of title group 28063, There are many Cabin info is missing The Cabin is related to Pclass We drop this feature no problem so far There are 2 entries of Embarked missing We fill it with the most repeated value S Age of many people is missing Again the simplest way to impute the age would be to fill by the average We choose median for fare imputation We use Spark s fillna method to do that For age we use more complex imputation method discussed below For now I am just focusing on the train data There can be different feature missing in the test data Acutally there is missed fair in test data we calculate median fair also We come to the test data at the end of this notebook 20735, HeatingQC column 2826, Xgboost 27896, Compiling the Keras Model 21798, Naive Bayes Gaussian 34457, TX 2 23445, First we have to separate the individual date and time for each data point into hour day month and year 43371, Splitting Data for training and Validation 25427, Merge datasets metadata 23543, Data Visualization 29706, Show a sample 4924, Define the Models and Do a Scoring 31245, Prediction 36432, Scoring 33799, Exterior Sources 271, Library and Data 13257, sibsp and parch 39438, Model Prediction 1148, we need to handle missing data 11997, On applying grid search cv we get best value of alpha 1 41672, let s take a look at some malignant tumours from the train set 15237, Handling missing values outliers mistakes 2078, You can check yourself the public LB score of this model by submitting the file submission2 3498, Training and estimated test accuracy 16602, Numerical Variables 961, calculate the mean of all the feature importances and store it as a new column in the feature importance dataframe 42945, Evaluating the cross validation AUC score 18925, Library 34851, Outliers Detection for Top Important Variables 25996, Querying the Data 11447, Encode categorical feature columns 39821, There are two call back techniques that I tried 7662, correlation train train 38418, CNN Keras 3519, Heatmap 41020, this is the total number of passengers in the test set by WCSurvived value we are finally ready to make predictions 28925, We use the cardinality of each variable to decide how large to make its embeddings 24369, Living area in square meters 34178, Saving the model as an hdf5 file so that we don t have to re train the same model 15157, Seperating the train and test data from the concatenated dataframe 7089, We haven t gererate any feature from Parch Pclass SibSp Title so let s do this by using pivot table 29853, we specify n estimators 100 28415, Visualization of a single decision tree 9077, it looks like most rows have values for both the Exterior1st and Exterior2nd only 1 null In addition it looks like most houses are made of VinylSd material Since we are feature engineering these 2 columns into multiple True False columns of whether the house is made of the material True False we don t need to fill the rows with Null since they be false for everything do it 37651, Create Model 12367, Electrical 38732, Since all of the passengers have the same ticket number we can conclude that the fare was calculated for the entire group and not each individual 11236, let s first use RandomForest to train and predict 11311, Null Value Calculation 13162, Model Selection SVC 14012, Analyze by describing data 10818, start from round Age 9161, OverallQual 1878, Gender 19141, Model 2 with GradientDescentOptimmizer 34750, Building model without pretrained embeddings 20437, Go to top font 14535, Logistic Regression 28450, CHECKING DATATYPE OF EACH COLUMN 5243, Transforming The Skewed Features 9203, Family Member Count 15633, Survival by Deck and Gender 19360, Distribution of numerical attributes 19627, Rather fitting dataset individually to every ML algorithm Pycaret gives feature to compare between them directly 28712, Importing the Datasets 22633, Prepare final dataset for Modeling 34152, Data wrangling 9240, Whisker plot overallqual saleprice 20469, Days from birth distribution 14919, SibSp Parch 763, Train the classifier 12628, Cross Validation 10994, Import data and explore it br 33804, There are 35 features with individual features raised to powers up to degree 3 and interaction terms 23241, Mean Absolute Error achieved is 1278 22696, Lets Try to Predict Other 28403, One Hot Encoder 2504, Age band 13884, Passengers Class 21023, Replacing words in a text document 11100, Percentage of null valued features in Train data 23691, Load data 6874, Combining Data Sets 12704, I stick with these bins for now 823, Conclusion from EDA on categorical columns 19166, SVD on tf idf of unigram of product name features 7689, Preprocessing 27359, I remove the negative values as they cause noise on the data 4997, Not much action in the 80s apparently 14243, Exploratary Data Analysis 32617, Exploring the location column 23794, When I started my journey of becoming a data scientist and honing my skills on kaggle as embarrassing as it was I could not find where the data is 37112, Code in python 23529, Training and Evaluating 18910, Embarked feature 22918, Among the top 50 mispredicted tweets only 4 are false positive 42404, Preparing the submission file 18011, Training set 18645, RENTA 5445, Split into X and y variables 10355, Hyper parameter Tuning 24123, Gerando Sa da 25726, make a directory for our datasets after unzipping 18000, The correlation matrix measures the linear dependence of the features and it is desirable to have features that have little or no depedence on each other 20248, It s time to use One Hot Encoding 24164, Top Mask Position 36482, Since we know the files that we be loading ahead of time we be making a list of some basic information about them then iterating through that list 39247, Format dataframes 31071, WARNING The Implementation of this cell takes resources 16781, Support Vector Classifier 14406, Map values in test set also 20615, Import libraries and functions for cross validation and metrics for accuracy 10986, The Last Step is to save the file 26691, BIRTH EMPLOTED INTERVEL the days between born and employed 42027, Spliting the strings per alphabet 30823, Full overlapping 32096, How to stack two arrays horizontally 13106, Logistic Regression 42291, Submit To Kaggle 2526, Bagged DecisionTree 42771, One Hot 5908, u ANN u 35850, Lets have a look at our data 17276, Decision Tree 7865, Supposing that the categorical values such as the Name the Cabin the Ticket code and the ID doesnt have any relationship to the fact that the passanger died or survived 9257, Random Forest Regression 10419, SHAP Values for one house 15868, Train model 5749, The algorithm decided that the attributes sex age and fare are the most important and were decisive 25268, Image Data Generator 8350, As noticed already before the class 1 passangers had a higher survival rate 29880, Without KFold 18322, Train and Validate Different Models 41810, Visualize model performance 20478, Credit status 35085, GridSearch Cross Validation 23481, Stopword 15396, There are 2 missing values for the port that the passenger embarked on in training data 12842, Decision tree 42101, Plotting an example of Image data from the training dataset 4051, Embarked Analysis 40650, XGB with hyperparameter tuning 24830, PCA 35222, Since we have 1 NULL row we remove it from train data 20724, HouseStyle column 31111, there is no obvious increasing or decreasing trend 6614, Split the Train and Test set from df final 5581, Data Cleaning 10961, MSZoning Utilities Exteriors Electrical Functional Utilities and SaleType 23555, It looks like people start preferring larger houses 5351, Diplay relationship between 3 variables 39130, Labels Hot Encoding 2163, At this point our model 7252, Pre process data 18239, Backpropagation Chain rule 982, Interpretation 30875, Plotting 2018, For our models we are going to use lasso elastic net kernel ridge gradient boosting XGBoost and LightGBM regression 9056, FullBath 36206, Using the facets layer to compare all the curves 23429, we do a bigram analysis over the tweets 3243, Datatypes and its distribution 4060, Which variables are most correlated to the age We can check the correlation table to get the answer 1814, Fitting model simple approach 32424, How the schema of this example looks 19344, Using a CNN 42520, Linear Model 16647, Importing Libraries 21417, NaN imputation be skipped in this tutorial 39098, The Coleman Liau Index 2480, Using another K 28611, We have 3 classes with high frequency however we have 3 of low frequency 13095, Frequency distribution Continuous Variables 25442, Confusion Matrix 4905, Looks like our Data is Skewed Towards Right 16720, sex 12199, Short note the custom dummifier 39921, Before cleaning the data we zoom at the features with missing values those missing values won t be treated equally 14517, Sex 12406, LotArea 12554, Model Evaluation 9337, But we can decide for example to half the number of bins to map our feature to obtaining 26540, Submission uncomment the lines below to submit 1651, A clean categorical feature here with 3 categories 35249, How can we utilize jaccard metric here 11495, Since Utilities feature have almost same value so we better remove it 33735, There are two common situations where one might want to modify one of the available models in torchvision modelzoo 16674, Making A Submission 13658, Embarked 2034, Age 42866, Prepare a KerasTuner search space 23185, Findings 19263, MLP for Time Series Forecasting 6246, we have missing values for Age Cabin Embarked and Fare 49, Correlation of features with target 16626, Feature Selection 9328, While if we correct for the mentioned sparsity we get 34096, Symptoms 7527, Plots 10202, Lets fill the null values with the mode of the respective columns 41739, that the embeddings have been trained lets check on how they were moved and if the different POS versions were disambiguated 6759, Checking Skewness for feature LowQualFinSF 23664, modeling 11563, let s check whether the resuduals are normally distributed around 0 19283, Make predictions 5128, Feature Importance 40300, Number of characters distribution as well is right skewed 39864, GrLivArea GarageArea TotalBsmtSF 1stFlrSF YearBuilt YearRemodAdd Numerical columns 19770, Train test partition 4199, For some variables it s difficult to know directly if there are ordinal or non ordinal 11909, Lets create a new feature column by combining sibling spouse parent children column 27843, Plot loss and accuracy 9003, Continuous Variables Distribution 22553, 0s 43028, Not that many mistakes 36342, Implement Forward Propagation 12318, Finding Missing values 6668, Feature Selection using RFE Recursive Feature Elimination 20913, Testing model 6196, Support Vector Machine using Polynomial kernel 28472, create a feature called season to know in which season transactions are high 32659, Once all numerical features have been preprocessed it is important to verify the correlation between each numerical feature and the dependent variable as well as correlation among numerical features leading to undesired colinearity 9170, Much better We leave this variable transformed 82, We can use the average of the fare column We can use pythons groupby function to get the mean fare of each cabin letter 5491, We can use Imputer library to take care of missing value but in this scenario only one value is missing in both columns so we update that with most frequent value and mean value in Garage Cars and TotalBsmtSF respectively 38574, Model Time 898, for DecisionTreeClassifier 22945, We have an easier feature to handle 35171, Experiment Replacement of large kernel layers by two smaller ones 10104, we have to get a Test data in Dataframe 7421, Remember that we have 76 variables for training and 75 for test before 11849, Modelling 30585, Putting the Functions Together 36139, One way is of course to tune the hyperparameters by hand 1208, Ensemble 1 Stacking Generalization 2881, Reading in the Test data 15564, to set them all on starboard as they probably gathered They traveled with ticket PC 17608 so 4412, Replacing missing values 21947, ML part Random forest 11975, Fill GarageYrBlt and LotFrontage 12072, Feature Engineering 2982, Correlation values 40862, One Hot Encoding 4879, Examine Dataset 25959, Top Selling Products 41161, FEATURE 7 2196, RandomForestRegression 31411, Updated Using resnet gives a boost of performance to LB score 3409, Create new variable Deck 21896, Utility Function 6426, From Scree plot we can conclude that we 60 PCs can explain around 90 variation of the dataset 3142, Few improvements using scalers and feature generators 28524, Fireplaces 39319, ASSEMBLE PREDICTION 11077, Feature Importances 8268, Create TotalSF Feature 38694, After Encoding 7211, lets plot a distribution plot to get a better idea of how our SalePrices are distributed 36889, reshape for CNN 18766, Due to computational limitations the size of each cannot very large too 26952, Submission Dataset 32553, Union Clusters 1124, Now apply the same changes to the test data 23710, Use the next code cell to one hot encode the data in X train and X valid 11653, XGBoost 19361, Find Outliers 38312, Random Forest 4217, Data Visualization 20107, Item count mean by month main item category shop for 1 lag 40727, Confusion Matrix 7391, the 5 unmatched passengers from kagg rest6 are added to the rest of the matched passengers in merg all2 20082, Top Sales Shop 15287, Creating categories based on Embarkment location 12898, Here we get a view of the dataset that we use in our predictions 12100, Dealing with missing values left 30353, let s display predictions for future weeks 39248, Export data 10974, Linear regression L2 regularisation 35865, Normalisation 5911, u XgBoost u 25408, METRICS 43381, Attack methods 14247, Age Continues Fetures 19130, Rounding data 847, RandomForestRegressor 27256, Extract feature importances for our Second Level 21242, The Discriminator model 29141, Binary features inspection 37704, let s look at our neurons 14935, Prediction on test dataset 1130, Exploration of Embarked Port 19309, Evaluation prediction and analysis 20179, Fitting Training Data 15330, do the same thing that we did with cabin so that we are left with the initials and can assign them numeric values accordingly 38454, But we can t use ent type directly 34915, Count words by whitespaces 28149, we use the networkx library to create a network from this dataframe 37508, That gave a score of 18667 12924, Sex 33770, The Pixel Values are often stored as Integer Numbers in the range 0 to 255 the range that a single 8 bit byte can offer 40278, Setting up the environment 3399, More Feature Engineering 25943, CatBoost 2658, Train a machine learning model 16749, PClass 30003, Submission File Preparation 31633, Lesson Learned 31804, Training the model on GPU 39036, But how many of them there are 6934, Categorical features 31745, ColorJitter 12836, Predictive Modeling 34371, Mean Absolute Error 24 29526, XGBoost 32308, As discussed before I have now categorized the Age feature into following two categories 17274, Linear SVC 25214, 99 of Pool Quality Data is missing In the case of PoolQC the column refers to Pool Quality Pool quality is NaN when PoolArea is 0 or there is no pool 41258, Fit The Model 7411, HouseStyle Locations matter a lot when considering house prices then what about the characteristics of house itself Popular house styles are Story and Story story and story nd level finished houses can be sold at relatively higher prices around dollars while the prices of story nd level unfinished houses are mostly around dollars Notably for multiple story houses level finished or unfinished have an obvious relationship with house prices 10794, double check it 19262, Train validation split 17825, Model with Sex Age Pclass Fare Parch SibSp features 27046, Patient Overlap 10445, Since GrLivArea is now normally distributed we shall look into TotalBsmtSF 32636, BONUS stacking 38096, Handling Missing Values 16397, Checking the correlation between attributes and Survived 31409, Start training using standard fastai 32118, How to find the percentile scores of a numpy array 15409, The median values are close to the means 9519, Define Training and Testing datasets 14622, Station 5 Categorical features 6904, Another way to present it 36065, Predict All Months 15322, We be searching for the initials of the cabin numbers like A B C etc 23730, Maximum passengers boarded from Port S while the least boarded from Port Q 7275, Sex Feature 32747, Comparison of 9 models including 8 new models 11858, Cross Validation Scores 9640, Most Correlated features 3413, It looks like traveling with 1 3 family members could have positively affected the probability of survival 7914, Check the missing values 33086, Dummy transformation 14967, Clearly survival chances of males is very low when compared to females 32536, Processing the Predictions 3414, FamilySizes of 2 4 are associated with a greater than 50 chance of survival per the sample 32582, The Trials object hold everything returned from the objective function in the 32880, Random forest 11056, Sex versus survival 17379, Fill Missing values in testing data 42861, Model evaluation 28206, Examine the class label imbalance 24870, that we have finished preparing the data we are ready to split it into a train and validation set using train test split from sklearn 21662, Filter a df by multiple conditions isin and inverse using 16940, Random forest 34721, Training the LGBM models on 5 separate folds and using their average prediction for the final submission 1913, This is not great we try some polynomial expressions like squareroot 32337, Tax amount 22095, Train our Model with simultaneous Validation 40185, Using my notebook 7552, Extra Tree 10576, After handling missing values we do some simple feature engineering 1595, Fare 32852, To mimic the real behavior of the data we have to create the missing records from the loaded dataset so for each month we need to create the missing records for each shop and item since we don t have data for them I ll replace them with 13982, Correlation between columns 31405, from df pass path filename to get image file we need to properly open the img when fn is passed 36608, How did our model do 41008, We now assign the label noGroup to every man and count the frequency for each Group id element of the dataframe in the new WC count column 16862, this is a classification problem 33731, Splitting the dimension of box in the formate xmin ymin w h 12134, Training the model 2 1859, Linear 23557, Seems there are too small houses 7007, Porch areas in square feet 14357, Distribution of Classes of Non Survived passengers 9092, I am also interested in comparing PUD homes verses not 14174, Before deleting the columns with prefix Deck and the AgeGroup column I check the accuracy score with and without these columns in 3 classifiers as a test to make sure removing them is beneficial 487, Correlation in Data 30676, I have preprocessed the text and just load it to increase re iteration time 1954, Creating Dummy Variables 21141, All the data analysis be done only with use of train data 21510, Try adding some gaussian noise 688, Scaling the numerical features below is important for convergence in some machine learning algorithms 7916, Check remaining missing values 6384, are used to visualize the main statistical features of the data mean value mode and 36260, looking at some satistical data 8698, In this section of the notebook I have handled the missing values in the columns 25037, Seems Satuday evenings and Sunday mornings are the prime time for orders 32264, Relationship between variables with respective to time 3694, Here we first use the numpy module namely the mean function 5606, Find 13175, First let s start visualizing missing values percentage proportion in each variable 6583, Fare per Person 40918, Functions to deal with Almon Lags 32996, Compare Ordinal and PCA 39243, Analysis of item categories 16752, Embarked 16392, Creating Submission File 33856, Distribution of the token sort ratio 2946, Drop features with with correlation value more than 38900, Imputing Missing Values 40804, Data Analyze by pivoting features 34759, Example 25859, Baseline Model Naive Bayes 12430, Without regex and using apply 12116, Predictions for submission 7725, Imputing LotFrontage with median values 15544, Time to train the test data 17919, Exploring the data further 8152, Lasso Model 38647, Age 20756, MiscVal column 23431, Removing urls 19146, Handling null values 10741, Use heatmap to check the correlation between all numeric variables 8959, Training model on training set 27432, Dropping first e 24233, Saving the model in Keras is simple as this 5462, Define our Metric 18098, After preprocessing we have managed to enhance the distinctive features in the images 35402, Training dataset 18949, Relationship between variables with respective to time with custom date range 25730, use torchvision datasets 39675, Define the optimizer to use giving it a learning rate and specifying what loss function it should minimize 19006, Tune the weights of unbalanced classes 19148, Modeling 20476, Go to top font 9802, Stack Model 2 After Manual Multicollinearity Check 32153, How to convert numpy s datetime64 object to datetime s datetime object 13408, Classification Accuracy 21401, Number of Kfolds 15438, And finally the Gradient Boosting Classifier 14387, Create a combined group of both datasets 22234, Confusion Matrix ile tahminlerin do rulu u Prediction verification with Confusion Matrix 26672, House Type 219, Library and Data 38545, dot Tpng tree dot o tree png 33235, Feature Extraction using pre trained models resnet50 9414, Feature eng Bins font 16375, Exploring Embarked vs Survival 1347, We can also create an artificial feature combining Pclass and Age 25910, Model Evaluation StratifiedKFold 39009, now write out the submission csv file 571, First Voting 71, Datasets in the real world are often messy However this dataset is almost clean 15162, Data Formatting e Discretization Datatype Coversion 28318, identifying the missing value in bureau balance 43119, make predictions 40411, Latitude Longitude 11768, Exploratory Data Analysis 18527, Training and validation curves 5682, Check if there are any unexpected values 3546, what if we use this function to visualize more precisely what s happening 24942, Q Q plot after MinMaxScaler 27240, Confirmed Cases 29735, The validation AUC for parameters is 0 36253, Extracting Features 9365, Predict Survived with Kears based on wrangled input data 18403, Trigrams 39747, Logistic Regression 33075, The two dots on the right of the graphs might be some outliers 4575, workout the numerical features 12645, k Nearest Neighbours 14383, Feature Age 1886, This IsAlone feature also may work well with the data we re dealing with telling us whether the passenger was along or not on the ship 15588, Well 40858, Since thse variables are highly associated with SalePrice mean SalePrice should be different across the classes of these categorical variables 17572, K Nearest Neighbours 37620, Numerical Variables 37347, Compile Your Model Fit The Model 27571, Well 23574, Optimizing Neural networks through KerasTuner 27213, I ll do some very basic preprocessing like 32109, How to pretty print a numpy array by suppressing the scientific notation like 1e10 29367, KERNEL SVM 3585, Changing OverallCond MSSubClass into a categorical variable 21408, Show AUC performace of best pairs of features 20394, Bernoulli Naive Bayes Model 19322, Output Visualizations 9482, Decision Boundary 37404, Read in Full Dataset 18136, If you can afford it 10 folds and 5 runs per fold would be my recommendation Be warned that it may take a day or two even if you have a GPU 17571, Random Tree Classifier 39313, Import and process training set 36427, Dummy Encoding 12932, Correlation Heatmap 5144, Setting Seed 21387, Model evaluation 8338, ee how the important features are related to our target SalePrice 4234, let s try the same but using data with PCA applied 4464, Name Title mapping 6879, Reading in data from the competition page 29724, Submission 34527, Seed Features 7266, I thought if free passengers could have ship personnel 14183, Cleaning the dataset 9371, BsmtQual Evaluates the height of the basement 7772, Voting 38631, That s it make the submission 2440, Trying out keras 32309, Find relation among different features and survival 33816, Model Interpretation Feature Importances 1051, Last thing to do before Machine Learning is to log transform the target as well as we did with the skewed features 27162, SaleType Type of sale 25279, We have to try and get the dataset into a folder format from the existing format which make it easier to use fastai s functions 35837, First create training data 42843, Germany 27754, Removing URL s 33666, Not sure about timezones 3392, Simple Tutorial for Seaborn 40313, Data Profiling 10382, the test set also needs attention 14412, I ll use Kfold cross vaidation to evaluate the model scoring method wil be acuracy so lets initialize Kfolds first 6222, for numerical columns filling NaN as median value 318, Prepare for model 12927, Embarked 36836, Train the Neural Network 28996, Features with multicollinearity 6077, Fare 7350, We need to adapt the test dataset in order to be used by our model 16680, There are null values in the age cabin and embarked section in the training data and in the fare as well in the testing data 40949, We now feed the training and test data into our 4 base regressors and use the Out of Fold prediction function we defined earlier to generate our first level predictions 6130, Missing sale type 4987, Fare analysis 13596, Final Predictions 32010, So in Ticket column there are 681 different entries in 891 rows This situation is named as High Cardinality We can drop Ticket column from both train X and test 18087, The most green images 1527, Pclass Feature 39116, RandomForestClassifier 12529, Encoding Our Data 5137, Continous Variables 31599, EDA ON THE CLASS DISTRIBUTION OF THE TARGET VARIABLE 42618, Training the network 15471, Correlation of Categorical Features with Survived 3876, This is unexpected why does building a garage after the house is built reduce it s price I know that this assumption is a stretch but let me know if there a reason behind it 8292, Fit model 43116, build model on oh train oh valid 9075, There are a lot I next wondered if the order of the covering types mattered 38722, so now we are going to create the generator for the DCGAN 38569, Variance Threshold 4478, We first sort out our training and test set 23003, Compared to other states TX stores have similar tendency regarding registered entries 32518, Compiling the Model 12695, Well that s a fare from ideal view There is quite a severe left side skew which probably won t pair up all that well with Machine Learning algorithms 19393, Train Word2Vec model 37143, Inference Code 23892, Bedroom count 5685, Ticket 509, Decision Tree 40842, Importing Packages and Collecting Data 7800, Stack Models 4098, Interpolation for Age 18830, add the previous averaged models here 2050, KNeighbors Model 9310, Ordinal variables and measurement errors 26459, GB Predict training data for further evaluation 11993, let s check the R squared value which is a percentage measure of variance explained by model 20853, We re ready to put together our models 42010, Sorting with conditions and get the percentage 31085, WORKING WITH TEST DATA 14815, Pclass Survived Age 20085, Monthly Aggregation 30873, The confusion matrix can be very helpful to evaluate the model 5854, Following the direction of the author we look at the plot of SalePrice vs GrLivArea identify the outliers with high leverage and remove the ones with 4 000 square feet of GrLivArea from the training set 27234, Create pipeline 24841, Simple CNN 7538, lets choose our feature attributes Name is not giving us any proper info so lets drop it Cabin column have various missing values and filling it may affect our prediction so drop it to Ticket also not needed so drop it 16899, Evidence proves Master Boy 16540, The Name feature is not really useful but we can use the Title of a person as a feature so let s do it 21601, Fixing SettingWithCopyWarning when creating a new columns 18549, AGE 20611, Fill the missing value by grouping by Pclass since cabins are related to class of booking 16346, Check correlation with Survived 8095, Sex vs Survival 15920, Data Transformation 15317, Checking out Embarked Attribute 5689, Divide the Train and Test data 29519, font size 3 style font family Futura let s have important data from Dataframe 21439, Bathroom Count 40021, Perhaps we can do both exploring the images and building up datasets and dataloaders for modelling 4785, pair the 5 most important variables according to our matrix with sale price 32425, Training on all Images 22602, Special features 11144, use the value as the improvement below this value is minimal 20235, I extracted only first letters of the tickets because I thought that they would indicate the ticket type 30647, Grid Search for random forest 18017, Family size new feature 29862, data element 10368, Numerical Values 38489, Predicted images font 13333, let s split the Name feature in order to extract the Titles and create a new column Title filled with them 7279, Family Size 11546, Visualizing Categorical Variables 5161, Decision Tree Regressor 29538, we re gonna reshape our image to let the model know that we re dealing with a greyscale image hence 1 color channel 10616, Fare is strongly correlated with Pclass 4313, Passenger Class PClass 25445, Testing Dataset 1057, Lasso regression 24807, Outliers 11282, In an earlier step we manually used the logit coefficients to select the most relevant features An alternate method is to use one of scikit learn s inbuilt feature selection classes We be using the feature selection RFECV class which performs recursive feature elimination with cross validation 37633, Instead of writing all of our code for training and validation in one cell it can be helpful to break the different parts into functions 33283, Family Assembler 13340, we are using the get dummies function to convert the Ticket Letter column into dummy columns that can be used by our future models 12507, Finished Submit Your Tuned Model 30870, Compiling the model 41931, We ll also create a device which can be used to move the data and models to a GPU if one is available 10684, Train test split 31607, feature X contains 8 pixels 784 28 28 15596, Statistical Overview of the data 27076, Part of Speech Tagging for questions Corpus 9514, Fare 32373, Getting Image Attributes 12561, let s ask some questions and get to know basic trends from data 16357, Choose and submit test predictions 14775, Survival by Pclass Socio economic status 13465, Considering the survival rate of passengers under 16 I ll also include another categorical variable in my dataset Minor 6396, Sales Price Analysis 35512, In this part I would like to create new features 11123, FEATURE ENGINEERING IN COMBINED TRAIN AND TEST DATA 6024, Understand the Target SalePrice distribution 6844, identifying the missing values 5513, Logistic regression 40264, Total Rooms Above Ground 43275, Avaliando o desempenho do nosso modelo nos dados de treino 39091, Question length 10946, 1again check size of data sets 40934, Tabular Modeling 2292, Making a pandas dataframe from either a list or a dictionary 29954, Price outliers are generated by some specific brands 6811, Neural networks are more complex and more powerful algorithm than standars machine learning it belongs to deep learning models To build a neural network we are going to use Keras Keras is a high level API for tensorflow which is a tensor manipulation framework made by google Keras allows you to build neural networks by assembling blocks which are the layers of our neural network For more details here tutorial deep learning in python is a great keras tutorial 34352, Besides the embedding 3 fully connected layers 32152, How to find the index of n th repetition of an item in an array 8150, Decision Tree Regressor Model 28928, Test 14445, Update Fare with ordinal values from the FareBand table 10858, Creating new features as per our intuition and dropping the other columns 5618, Code Output 2216, Blend 13255, Age 4847, Lasso 37005, let s combine them in a single dataframe 4058, it s necessary to normalize the data 37092, Statistical Functions 9368, Build model and predit 36571, Fine tune 8362, Pclass because there is only one missing value in Fare we fill it with a median of the corresponding Pclass 14623, If we would predict a probability of for a male and for a female to survive we would expect gradient contributions that point into opposite directions But actually we obtain 7424, To address the multicolliearity problem I apply PCA to decrease the number of variables 18410, Metric 26562, take another look at the scatter plot 27297, Check the total US trend 32203, Create a test set for month 34 8672, Torch model 28399, PairPlot 15457, Title 30394, Oversampling 32684, Cross Validation 30459, Split datas in train and test set 17570, Decision Tree Classifier 39109, Fare 6992, Original construction date 3844, option1 replace all missing age values with mean 30370, Test PyVips 32383, Loading Modelling Tools 26879, Score for A1 17688 9935, Therefore the missing age values are handled p 18184, Blending 28420, Item id 13461, I ll also create categorical variables for Passenger Class Gender and Port Embarked 212, Library and Data 41582, While using transfer learning in ConvNet we have basically have 3 main approaches 13151, apply title wise age filling in the transformations back in test data too 6727, YearBuilt Vs SalePrice 29581, as standard we ll load the dataset with our transforms 25022, Anyway when you want to use this mask remember to first apply a dilation morphological operation on it e with a circular kernel This expands the mask in all directions The air structures in the lung alone not contain all nodules in particular it miss those that are stuck to the side of the lung where they often appear expand the mask a little 1566, Fare 21513, Lets checkout the message length whether the sms is a spam or not 42046, factorize 1111, Evaluation 23062, we need to create few additional parameters for our model 7987, Linearing And Removing Outliers 37066, Impute Age 8122, Perceptron 10830, it is time to update ticket number with new category 32511, Pre processing the predictions 3786, Correlation Matrix of SalePrice 3536, PointPlot 26259, Training the Neural Network 26262, Reading test file 3161, The next step is to transform numeric variables to produce better distributed data 17743, Unfortunately it does not look like tickets were issued in this manner 10270, Neigborhood There are quite a few neighborhoods Surely these can be reduced to a few classes 4951, Modeling 32567, Diff Common marking 24177, Things look good use that to train on the whole data set 43211, Model training visualization 7032, Type of Paved driveway 42248, Transforming data to reduce skew 33552, How to win 40055, let s try to find good learning rate boundaries 35053, Train the model 26406, The survival chance of a passenger with 1 or 2 siblings spouses is significantly higher than than for a single passenger or a passenger with 3 or more siblings spouses 42134, Quadratic Weighted Kappa 13280, Support Vector Machines are supervised learning models with associated learning algorithms that analyze data used for classification and regression analysis Given a set of training samples each marked as belonging to one or the other of two categories an SVM training algorithm builds a model that assigns new test samples to one category or the other making it a non probabilistic binary linear classifier Reference Wikipedia 42097, Dividing the train data set into dependent labels and independent pixels features 22233, Periyotlardaki do rulama kay plar Validation loss in epochs 32165, Generate test predictions 15519, we can look at Ticket again 36194, The date in which the customer became as the first holder of a contract in the bank can not help fix the issue with missing data in the new customer index 43043, Any help on this is appreciated 35826, To learn about boxplot you may follow the links 35480, ORB Oriented FAST and Rotated BRIEF 25787, there are 2 missing values here so let s replace it with most frequent value 35818, Add sales for the last three months for similar item item with id item id 1 9651, Missing Data Percentage Visualization for Clarity 12104, Evaluation 10525, We have removed co linearity from our dataset manually examine each feature and remove non linear features from the dataset 17014, PassengerID 27875, Average daily sales 11683, Feature Engineering 8147, Defining Training Test Sets 4270, These where Sal but in order to have the same structure I put them in Sev 541, New Feature Family size 21626, One hot encoding get dummies 21843, RNN Architecture for MNIST Classification 28865, Attention 7478, 3d ParCh 7628, list of models 15574, Update After trying a few classifiers I realize that this really happens The remedy could be to subtract one form TicketGroupSurvivors where a singleton survived 26068, Another experiment we can do is try and generate fake digits 23598, I was first going for merging the two sentence sets into one so that the system is trained over all the words sentences document vectors and I would not be able to do it without concatenating both of these 43347, We fit our model to X train and y train datasets to train our model 15830, KNN Classifier 9879, Correlation Between Parch Survived 7479, 3e Fare 9145, Great Since there are only 2 and 3 missing values in these columns respectively and their values have the same range I just set the null values to the value of the other 17688, AGE SURVIVAL 34289, Classifier 37799, Model Evaluation 9700, Concating numeric and categorical features 18500, Kaggle Submission 79, combine train and test data first and for now assign all the null values as N 16282, Feature Importance 12697, The full names as they are not be helpful to us although there s probably something useful within title e 679, The easiest method to combine different classifiers is through a Voting Classifier It does exactly what the name suggests each individual classifier makes a certain prediction and then the majority vote is used for each row This majority process can either give all individual votes the same importance or assign different weights to make some classifiers have more impact than others 6733, GarageArea Vs SalePrice 16122, Perceptron 1109, Logistic Regression 11245, Multicolinearity 31049, Get ID 7298, The plot of SalePrice is skewed in nature 11465, Support Vector Machine SVM 38989, Traing for negative sentiment 2107, This residual plot is fairly problematic and we address it later on 19626, Loading and normalizing the dataset into Pycaret 4554, Utilities For this categorical feature all records are AllPub except for one NoSeWa and 2 NA 3746, LonFrontage 12303, Random Forest 33655, Define Gini Metric 35337, Fitting on the Training set and making predcitons on the Validation set 27601, Submission 11223, Show both adjustments 6873, Categorical Nominal 31107, One Hot encoding a Column ith Sklearn 19702, check if we have NaN values in our dataframe 25774, Male is more than female in the Titanic 22912, Evaluate and Improve Our Model 17388, Dendrogram 28692, Target Variable 43127, Predicting actual test set 13460, According to the Kaggle data dictionary both SibSp and Parch relate to traveling with family 42000, Sorting a certain column with a certain condition 23121, Findings 21151, First let s prepare the data set containing predictions of our future models 19517, Checking Number of Partitions 14098, Support Vector Machine SVM 15103, We were able to fill in the gaps for Age and now have 1309 records 2243, Pipelines 2033, Class 7229, Miscellaneous features missing values 18947, Relationship between variables with respective to time 40429, Examine the variable importance of the metalearner algorithm in the ensemble 24188, For example here are some mistakes and their frequency 40871, Optimize Elastic Net 15168, We can do even more things with these values 40195, Split data into train and validation 9614, Box Plot 29993, Building a scoring function 40284, Focal Loss 12785, Define the Training as a Function 1248, take a look at the distribution of the SalePrice 8933, Label Encoding 21486, The difference is very significant and suprisingly so While spawning more threads than there are CPUs available isn t helpful and causes multiple threads to be multiplexed on a single CPU it is unclear why that overhead causes xgboost to perform slower by several multiples 26851, Bigram Insights 39084, Embeddings 8756, See some start data 38959, It s always strongly recommended to check your input pictures into your model after augmentations to ensure that not strange things happen 5314, Preparation for creating functions 16365, Sex Male Title Mr Pclass 3 lowest chances of survival 19321, Training the Model 2419, First let s take a look at our target 40800, Data Pre processing 21663, Reverse order of a df 1553, Pclass 43018, Feature assembly 35071, Below there is the definition of the standardizer function which preprocess each one of the images as they are fed to the CNN 39947, Cross validation 16362, Name could be one of the interesting features as it directly doesn t provide any value but it opens up the scope for feature engineering from it 9009, Set Fireplace Quality to 0 if there is no fireplace 24539, Total number of products by income groups 41762, Not surprisingly this search does not produce a great score because of 3 fold validation and limited parameter sampling 39194, Modelagem 37222, Interesting Very few vocabulary words are in common with the paragrams vocabulary 1008, How about having another look at our features 40980, review How to count rows in each group by using 15407, let s create a Married column and assign titles that indicate the woman is married the value 1 and 0 otherwise 38977, LSTM model with Pytorch utilizes GPU 27757, Removing Punctuation 11737, Here we are looking mostly linear again although it looks like it might be following a little bit of an exponential path 1429, I put here some hyper parameters tuning with n estmators max depth and learning rate parameters 4571, NA for PoolQC means No Pool 6500, all necessary features are present in both train and test 34265, Compare one example again to verify that the normalization was done properly 8963, data engineering 21485, try different submissions with different thresholds 3254, strong Number of Houses Sold every Month strong 5527, Create Pclass Categories 9852, XGBoost 35545, use pipeline to chain multiple estimators into one 7692, Linear regression 2827, optimizing hyperparameters of a random forest with the grid search 41369, Sale Price HLS Low Lvl Bnk 24658, Select only March and April 32724, Clipping some outlier prices 35930, SibSp Parch 22695, Binary Classifier s 8381, Looking for missing values 42254, Use scikit learn s function to grid search for the best combination of hyperparameters 25939, Below is a comparison of w o PCA and w PCA correlation after PCA transformation it looks much better in terms of high correlated variables 32680, The commented code below enables generating and saving a contribution to the competition 33669, Year font 33849, Analysis on few extracted features 43263, Importa a classe DecisionTreeRegressor do scikit learn 28024, Training Word2Vec 33444, ExtraTreesClassifier 9707, Ridge Regression 26101, Random Forest Cassifier 28460, First I would like to take all columns of type float in a list 13760, Model Data 7894, and drop the respective columns 8992, We expect that if the MasVnrArea is 0 that the MasVnrType is none 19353, Reading input file 12779, Submitting 27019, Training in progress Please do not disturb 35186, the highest correlation is among cont11 and cont12 32094, How to reshape an array 27644, that all of our data is ready we can go on to training the classifier 18130, get the RandomForestRegression model s assessment of the Top 5 most important features 26505, that the image size is reduced to 7x7 we add a fully connected layer with 1024 neurones to allow processing on the entire image each of the neurons of the fully connected layer is connected to all the activations outpus of the previous layer 40328, Locations 29108, To use FASTAI library we need to feed our data into their ImageDataBunch function However 20261, Steps of Linear Regression 29365, K NEAREST NEIGHBORS 3560, Run on test data 8661, Instructions 5166, Download datasets 37234, Evaluation 39134, Input Layer 23552, Submition 37921, Lasso 13427, Define custom Scorer function 23739, Correlation and Feature Importance 26316, Gradient Boosting Regression 4115, Exploratory Data Analysis 36756, A training set of 42000 images 28759, Stepwise Model Selection 34859, RF 39959, Averaging Regressors 2116, It looks like the target encoding with the mean is very powerful and not very correlated with the other features 25857, Splitting dataset 25278, To get a set of transforms with default values that work pretty well in a wide range of tasks it s often easiest to use get transforms 19553, CNN Model 39146, all folders are created The next step is to create the images inside of the folders from train 36584, Visualizing and Analysis 24887, convert non numeric features to categorical values 2804, get cat feats Returns the categorical features in a data set 13972, Embarked vs Pclass 19779, Let s take a look at the architecture 17339, XGB 22151, BIMODAL DISTRIBUTIONS 1578, because we are using scikit learn we must convert our categorical columns into dummy variables 22087, Normalization 30427, Load Models 7921, Combine features 25245, Prepare submission 6112, Ah what a pity mistake 11766, we are using a stacked model of our Elastic Net Kernel Ridge and Gradient Boosting Regressor with our meta model being our Lasso model in order to predict our outcome 579, Quantitative 1stFlrSF 2ndFlrSF 3SsnPorch BedroomAbvGr BsmtFinSF1 BsmtFinSF2 BsmtFullBath BsmtHalfBath BsmtUnfSF EnclosedPorch Fireplaces FullBath GarageArea GarageCars GarageYrBlt GrLivArea HalfBath KitchenAbvGr LotArea LotFrontage LowQualFinSF MSSubClass MasVnrArea MiscVal MoSold OpenPorchSF OverallCond OverallQual PoolArea ScreenPorch TotRmsAbvGrd TotalBsmtSF WoodDeckSF YearBuilt YearRemodAdd YrSold 15362, pclass A proxy for socio economic status 30591, Insert Computed Features into Training Data 30396, Callbacks 26389, Evaluation 6420, We can drop MSSubClass YrSold MoSold as they have no impact on SalePrice 39023, Machine Learning Data Analysis 16336, Linear SVM 35176, Compile 10 times and get statistics 22052, First orientation get some hints on what to look for in our EDA 3440, Since our Jonkheer is 38 years old let s include him in the list of titles we change to Mr 42815, Configuration 42141, Autoencoder Model 52, Random Forest 23336, BanglaLekha Isolated 19151, Write the data to file to save it for a new session 15610, Gender Feature 19977, MLP Sigmoid activation ADAM 32148, How to drop all missing values from a numpy array 3890, Range 11289, Replace missing values with most common value 3403, Normality 35421, Checking shape of training data and labels 15904, Exploring the Data Distributions 6478, Relationship with categorical features 35665, Features with alot of missing values 15748, Embarked is likely S for both 3723, Hyperparameter Tunning 38780, Express X axis of training set frequency distribution as logarithms and save standard deviation to help adjust frequencies for time trend 42356, Removing twitter handles 22772, We build the iterators 12602, EDA 32204, Concatenate train and test sets 12938, Deck Where exactly were passenger on the ship 43404, These are the parameters and their ranges that be used during optimization 15898, SHAP values for selected rows 42549, Question character length correlations by duplication label 17269, First import all required machine learning libraries 24309, accuracy vs 4591, We first convert GarageQual GarageCond and GarageFinish to ordinal numerical variables 11507, LASSO Regression 1416, TypeOfTicket vs Survived 24225, Sort the traning set Use 1300 images each of cats and dogs instead of all 25000 to speed up the learning process 6717, Relation between Continous numerical Features and Labels 25201, Evaluating our model 10405, Drop the Id column since it isn t worth keeping 1984, Great Our data is ready for our model 14802, Submission 22980, Promotion impact per store week 40385, According to the tensorflow website The tf 33144, To help the convergence of the map it is a good idea to limit the number of points on which the training be done 42218, A second Max Pooling Dropout layer now be added 4055, Defining Transformations 13824, now let s writ the model 10230, try XGBoost classifier model also 40655, I m happy to lose 5 of the features and not have to worry about a proper imputation strategy 21644, Split a string column into multiple columns 404, XGB 20788, From these regplots we have confirmed there are outliers so we decide to remove them 30566, We can test this function using the EXT SOURCE 3 variable which we according to a Random Forest and Gradient Boosting Machine 6817, There are lots of ways to deal with missing values 29102, LSTM Classic 27507, Compiling and Fitting the Model 7293, Random Forest 42550, Model starter 26670, Occupation Type 29699, lets take them through one of the kernels in next convolutional layer 12665, that we have trained the classifier and made some predictions for the cross validation set we can test the accuracy of the predictions using the following metrics recall precision f1 score and accuracy 15133, Replacing Rare value with one of clear titles 21070, Testing data comes in 3290, we create an object of this class and use it to fit and transform the train data 19581, Aggregate Sale 11544, EDA and Visualization 8087, Blend with Top Kernals submissions 39757, Here we ll try to create meta features to understand better the structure of the questions 34663, Item counts versus item prices 42609, Set Up Progress Tracking 35066, Complexity graph of Solution 4 1 407, Gaussian Process 24886, drop useless features 24357, investigate for errors 34329, Adding augmented data not improve the accuracy of the validation 14853, Family type 39456, checking missing data in bureau 29587, A solution to this is to renormalize the images so each pixel is between 34052, XGBOOST Training Cross Validation 12455, This means there s no garage 31547, replacing each missing values with the top most category 20064, Adding column item main sub category 37196, Skewed features 8891, LightGBM 1293, Prediction 5 5 27493, SUBMIT PREDICTIONS TO KAGGLE 36341, Define Neural Network Parameters 31081, UNIVARIATE SELECTION 13447, Data Cleaning 38008, All info about a single department 23806, Drop Low Info Variables 33357, Question 2 Create a decomposition plot for a city of weekly sales 18296, Data Augmentation Oversampling 41484, Random Forest Training 3471, For this model precision can be thought of as the ability of the model to pick out the true survivors and not label passengers as survivors when they in fact perished 3634, Pclass Survived Vs Age 17466, Age 40183, Long selected text are not predicted correctly 13112, Linear SVC 27463, Missing Values 449, Box Cox Transformation of highly skewed features 39408, Pclass 27316, Submission 3701, we have a model 33467, CompetitionDistance 27033, Use this mothed to predict test csv 31578, I could be wrong but in my opinion this is too much imbalanced 38045, Central Limit Theorem 33753, Numerical data processing 33617, Cleaning of test data for prediction purpose in the same manner as train data 7060, Scaling with RobustScaler 10787, I would like to check what would be the score 4569, SalePrice is not normally distributed 13472, Nearly all variables are significant at the 0 15563, Feature engineering 3968, LightGBM 36817, that we ve downloaded our datasets we can load them into pandas dataframes First for the test data 19706, Resizing the photos 17477, Voting Model 42142, simply visualize our graphs 13948, Categorical Variable 24847, As I was using an enriched dataset during the Week 2 competition I have to add the new countries to my dataframe and fill the missing data with median values 11397, Looks like these passengers share the same ticket information and were both singletons travelling first class 35143, Nice we got an impressing 98 32849, Data leakages 5809, Combine train and test 38406, Look at the digits 42249, Considering highly correlated features 42957, I think the idea here is that people with recorded cabin numbers are of higher socioeconomic class and thus more likely to survive 36141, First we have to define a dictionary of hyperparameters we would like to check 31727, Anatomy and Target 7035, Here we analyze correlation with the boxplots and missing values 24950, Dataset information Pandas Profiling 11137, Tansformation of the target variable 8430, Correct masonry veneer types 9349, Create a pipeline 38892, Forward Propagation with Dropout 38095, Missing Values 33465, SchoolHoliday 33749, Using the learned model to predict the test set 21079, The correlation coefficient for ps calc is 0 so we drop these from our dataset 42625, Fatalities percentage 25178, Featurizing text data 3295, SalePrice distribution 22404, Fill in missing with the more common status 2521, Confusion Matrix 8712, The Gradient Boosting gives the best performance on the validation set and so I am using it to make predictions to Kaggle on the test set 7992, Encoding Label 31060, Negative look ahead succeed if passed non consuming expression does not match against the forthcoming input 32107, How to create a 2D array containing random floats between 5 and 10 22782, Histogram Type 39820, it s all set now let s build our model shall we 31779, Loading datasets 4027, Interesting columns 15452, Change Survived to an integer 38967, Word2idx and embedding dicts for LSTM model 29805, Implementation of Glove via glove python 5077, How about our cleaned data and some other classifiers 155, Import libraries 32250, Visually inspecting our network against unlabeled data 1808, Dealing with Missing Values 12790, We thus export the predictions as a Comma Separated Values file 26668, installments payments 2809, Heatmap 5414, Except those with more than 4 family members basically the bigger the family size the more likely to survive 14097, Random forest output file 27135, We ll have a look at the correlation between all the features with the help of Heatmap 15861, Model definition 12975, We cannot make a prediction about survival condition by using passengers names but there might be relationship between survival rate and titles 21416, Label encoding Making it machine readable 42004, isin filtering by conditions 33783, Examine Missing Values 8803, replace the NaN in categorical with NA and with 0 in Numerical data 69, This is a sample of train and test dataset 19370, Missing null values in categorical columns 10923, To find the degree of a graph figure out all of the vertex degrees 20578, Encoding Categorical Data 36677, stop words string english list or None default 12133, XGBRegressor 2 22232, Modelin uygunlugu Fit the model 17409, Feature importances generated from the different classifiers 32339, latitude and longitude 26462, In our data formatting we generate the respective labels for dogs 1 and cats 0 for our training data File path also be collected as a column for our dataframe so that it can be used to load and train our images 4705, Features coefficients 5948, create a submission csv of prediction 15075, Age Group 22065, Findings for large spaCy vocab 42620, Import Libraries 9761, Lets check TitleGroup I saw that there are so many Mr 21227, Data augmentation 20061, Import libraries 10367, Missing Values 3451, For the sake of modeling let s add Mr 10426, Linear Regression 37686, Plot Predictions 36396, Features engineering FE 37798, Linear Regression 38781, Adjust frequencies for time trend 21436, make LotArea LotArea log 9826, Fare Passenger Fare 17602, Support Vector Machines 41941, save a batch of real images that we can use for visual comparision while looking at the generated images 7647, Generally if a variable is numerical we replace null with 0 or mean median 12827, same way we can check for unsurvied travellers family size 3583, LotFrontage Linear feet of street connected to property 31011, We have used categorical crossentropy as the cost function for that model but what does we mean by cost function 2319, Fitting a Linear Model 18314, Explore Items Dataset 2344, Gradient Boosting Classifier 32642, Warning PyCaret setup work for last imported PyCaret library 15710, Gender per Passenger Class vs Survived 24716, Eigenvector 4 15584, Only 102 9050, Garage Cars 21230, Run the following cell to visualize how many images we have in each of our datasets 7783, value counts in ascending order 21560, Binning 23882, check the dtypes of different types of variable 11138, Here is the difference between BoxCox and Log values the difference is substantial at the value of a house 32701, Here I am concating data to apply cleaning on both train and test set 5832, we have 18 object type feature 12 numeric 10948, again size of full data set 23247, We fill the Embarked s missing values with S I prefer it because it is the most frequently used 35438, The best part 33107, Submission 2263, Age 14103, Most of the children got survived 21656, Reduce memory usage of a df while importing duplicated Trick 83 32588, Continue Optimization 4955, ElasticNet regression model which basically combines Ridge regression and Lasso regression 27905, Simple Model Catch Failures Decision Trees 38515, Distribution of top unigrams 485, Defining a basic XGBoost model 6115, I think there should be a strong correlation between Garage Area and number of places for cars 35406, Applying desicion Tree 41323, The data consists of a tabular format where every pixel is a feature 3004, Ridge Regression 38549, The Meta Features Based On Word Character 24099, Limiting the outliers value 22948, visualize the Cabin Number distribution of the data 21918, Ridge model 42043, Inserting the average age per Initial for NaN values of Age 1915, MasVnrType and MasVnrArea 13738, Creating O P file 40066, Data Processing Reduce Skew on x 14891, Apply the pairs plot 22953, Show some image 26197, We use a different set of data augmentations for this dataset we also allow vertical flips since we don t expect vertical orientation of satellite images to change our classifications 34479, Perprocessing Data 16945, XGboost 14151, to examine the data types of our features 22518, Categorical encoding 25273, Display Validation Accuracy Loss 28122, Exploring Target Column 616, These are two women that travelled together in 1st class were 38 and 62 years old and had no family on board 32502, Generating csv file for submission 3680, Key Takeaways 25002, Run predictions for given test data and submit the output file in required format 25379, Evaluate the model 1705, Listwise Deletion Dropping rows 30908, We get only around 140 active features which are far away from the total features 1920, All categorical variables contains NAN whereas continuous ones have 0 19048, There are a number of Nan values within each column that we want to get rid off 27857, Handling categorial variables 15857, Create final feature set 18304, append test data to matrix with next month s date block num 1857, Lasso 22688, Prepare submission file 18526, For the data augmentation i choosed to 36352, Preprocess the Inputs 5521, Extract Title data 813, Filling missing values 7700, We try other kind of regressors such as XGBRegressor and ExtraTreesRegressor 32723, Calculating different aggregations 32720, Predicting on a random sentence 26843, DAE 15154, Filling Age null Values with random numbers generated between its mean and standard deviation 39100, Dale Chall Readability Score 33081, Another way to visualize this lack of data is to display the whole dataset as followed 28653, Neighborhood 17438, make some graphics by type 16272, Shape 10353, Principal Components Analysis 30347, Get data from 7793, reload the data so we can have a fresh start 21521, Lets visualize the Accuracy over the Epochs 41193, check if there are any numerical columns left with missing values 36266, Looking at Port of embarkation 16100, Embarked Feature 13998, Split train data into training data and validation data 20208, Standardize data 40466, Other Features 27327, it is to optimize a size of n estimators for RandomForest 3495, Parameters for the best model 40266, Total Basement Surface Area vs 1st Floor Surface Area 31684, Training Autoencoder 2995, Handling Missing Values 10124, Boys Girls and Women have higher chance of survival where the men have the lowest 13388, 0 indicates that person is travelling with family and 1 indicates that he is travelling alone 40038, Predict on whatever you like 26207, First we check if all images in the train folder are all in jpg format It is better to check because if there are a mixture of image type we may face troubles later on 29844, Transforming some numerical variables that are really categorical 20314, I think another interesting information my come by lookig at the 10th to 15th most bought products for each cluster which not include the generic products bought by anyone 32402, Submission 14840, Another interesting thing to look at is the relation between Age Pclass and Survived 7955, Tuning on input dropout 551, RFC features not scaled 35586, Data Imbalance 492, visualize missing values using seaborn 11022, Lets fill the two blanks in Embarkation attribute with S 36070, Prediction 20521, remove the outliers 3003, Ridge Regression 27970, Data set fields 39250, EXTEND DATAFRAME TO PRODUCT SHOP x ITEMS EVERY MONTH 29777, Required Imports 24972, Dataloader model 1991, use test dataset 12881, Parch and SibSp 10636, Start with 23463, Training set 21751, Weighted Averaging 12690, Fare 28235, Added one more layer with filter 128 42390, Price should be a good proxy for item id and gives it a numerical value instead of a categorical value 1928, MiscFeature 32626, TFIDF Features 41847, Moscow Price per square meter 17541, Find best parameters for XGBClassifier using GridSearchCV 38266, now we remove all the all the urls and the HTML tags 37415, Test Small Dataset 7454, Missing values in Cabin 21669, Submission 5165, Import libraries 946, Tune Model with Hyper Parameters 28500, Importing Libraries 32223, Seems like a pretty uniform distribution of all numbers 1 9 take a look at what the numbers themselves actually look like 2283, Test new paramters 26469, Lets fit our model with the training data and evaluate its performance 36703, True Positives TP True Negatives TN False Positives FP and False Negatives FN are the four different possible outcomes of a single prediction for a two class case with classes 1 positive and 0 negative 13448, now no missing value in Embarked 40975, size method counts rows in each group 12077, Modelling 13067, Decision Tree 33870, XGBoost Model 25651, Generate test predictions 11533, Discrete Variables 9943, Creating a feature which tells us that the person is travelling with someone or not according to the similar number on the tickets 19364, with reference to the target the top correlated attributes are 43062, check the distribution of the standard deviation of values per columns in the train and test datasets 15052, Same LastName doesn t make the same family but same family usually have same LastName 22145, Model definition and generating predictions 4758, Correlation coefficients 9680, Finalize Model 38833, Run inference on the test data 36573, Make predictions 33472, COVID 19 tendency in China 37432, Which are the most common words font 5306, The issue is that after your scaling step the labels are float valued which is not a valid label type you convert to int or str for the y train and y test to work 34166, we can translate the new dataset 32688, First we create generators for augmentation of training data and for normalization of validation data 10171, Count plot to comapre two feature 29549, Bigrams 16619, CABIN EMBARKED Age 27375, adding the lags to the test data 25947, Outlier Removal 14217, The Chart confirms 1st class more likely survivied than other classes 6515, As the continuous variables are all skewed we use logarithmic transformation to visualize 21953, Data augmentation 10951, Check the struture of full data set 15612, Title Feature 31624, the test set 13161, Unscaled Features 33319, Logistic Regression 35099, clicks train csv 9301, OLS experiment multicollinearity 16757, Missing data and Combine 15097, Neural Network 37172, The training error looks pretty good 41084, Albert Base 1125, Exploration of Age 7066, GridSearch for XGBoost 3846, option3 replace values with median Age of Pclass 2813, Saving the Id column for later use 27376, testing the lags impact 13605, When handle missing return nan Embarked column is replaced by two columns Embarked 1 and Embarked 2 25795, Age and Fare also include in Categorical data 26972, Define Train Model 26203, This is a serious problem that one can run into when you Normalize the bounding box it may exceed 1 and this cause an error especially if you decide to augment the images as well 27761, Removing stopwords 6201, Random Forest Classifier 33836, Pros 20264, Artificial Neural Network ANN 39076, Training 38034, take a look at these features and plot them on a box and whiskrers chart 21819, Correlations 39995, Submission 31248, Title 5547, Specify Architecture 32092, How to replace items that satisfy a condition with another value in numpy array 5709, KitchenQual Only one NA value and same as Electrical we set TA which is the most frequent for the missing value in KitchenQual 20790, Highly Correlated Features 39867, Feature Engineering 166, To be able to make a valid submission a good habit is to check the sample submission file provided by Kaggle to become familiar with the needed format 29818, Vector Averaging With pretrained Embeddings 10562, Merge Numeric and Categorical Datasets and Create Training and Testing Data 8333, let s have a look at the target distribution 26003, log loss or logarithmic loss is a classification metric based on probabilities where it quantifies the accuracy of a classifier by penalising false classifications in other words minimising the log loss means maximising the accuracy and lower log loss value makes better predictions 40409, Price 4986, Sex Age and Fare analysis 19094, Having more sibling can be corelated to less survival rate 12263, Data 21458, Split the Training dataset into Train and Valid 8944, Fixing Exterior1st Exterior2nd 37152, To find out the images that were wrongly predicted I have followed the following steps 2717, Kernel ridge regression 19439, Defining hyper parameters 14983, we can replace a few titles like there is Ms which can be replaced with Miss as they refer to the same gender group And few like Major Capt etc replace them with others 7260, Sex Feature 9208, Salutation Distribution 26548, Evaluate the model 32995, Train with PCA reduced data 12235, The residual sum of squares is the top term in the R2 metric 42950, Pclass Feature 34477, Removing id column to delete row use axis 0 column use axis 1 20800, Create TotalSF feature 1091, further explore the relationship between the features and survival of passengers 38814, Combinations of TTA 11986, plotting the correlation maps of some features say first 30 features on SalePrice 5121, Train and Validation sets split 25048, Looks like the products that are added to the cart initially are more likely to be reordered again compared to the ones added later This makes sense to me as well since we tend to first order all the products we used to buy frequently and then look out for the new products available 24845, Preparing the training data 38424, Convolutional networks 34605, Extract train features from CNN 17977, Name 31823, Random over sampling 21944, Spark 10251, Go to Contents Menu 27392, Tuning n estimators based on small learning rate 36053, Fill missing timestamps 39411, SibSp 42158, Merge train and test data 9323, For example thisi is how Foundation looks like 7883, Following the graphics below The age can be groupped into less classes 36110, Lets concat Train and Test datasets 13327, Age completing feature with its mean value div 33658, Set parameters 20921, model 19807, to encode categorical variable with k labels we need k 1 dummy variables 35629, so our digits are in a space in which every pixel is a variable or a feature 35475, Use Noise Reduction 36743, Here a dataframe is created to store the knowledge if an event exist in the next day 20026, Boxplot gives even better insights 22134, select only top 10 correlated features 42926, Dendrogram 19377, Cross validation 40116, seqstatd function returns numeric data 28527, WoodDeckSF 21517, Creating Our Model 36284, Wohh that s lot s of title So let s bundle them 37044, y test consists of class probabilities I now select the class with highest probability 10114, Again people with more number of children parents on board haven y survived neither did people who were travelling alone 8960, Creating output files 34351, There s no easy way of using the fast 27139, Category 1 Type of dwellings 9253, Linear Regression 28132, Will Implement It soon 19334, The training accuracy steadily increased and plateaued while validation accuracy is also consistent 30154, randomforest and gradientboostingregressor 15283, Survival by Age and Fare 42972, Reading data and basic stats 4646, Swarm Plot 1793, In probability theory and statistics Kurtosis is the measure of the tailedness of the probability distribution of a real valued random variable So In other words it is the measure of the extreme values outliers present in the distribution 2277, Which is the best Model 20635, lets apply this vocab on our train and test datasets we keep only those words in training and testing datasets which appear in the vocabulary 29458, Linear SVM 6706, Categorical Features 20856, We use the cardinality of each variable to decide how large to make its embeddings 30776, Train linear regression model and make prediction 18506, DICOMs 21011, Exploratory Data analysis with dabl 20617, Decision Tree Classifier 13885, Passengers Age 9105, Cool 33025, Compiling the model with the right Optimizer loss and metric 3904, Discover you Data what is look like 11710, Visualization 29982, Plot the cost and accuracy 35466, Visualiza the skin cancer Melanoma 32332, correlation map 15649, Gaussian Naive Bayes 9128, Lot frontage 2306, Python how to make a continuous variable categorical 2 ways 998, corr Fare Pclass is the highest correlation in absolute numbers for Fare so we ll use Pclass again to impute the missing values 13623, Before we start we need to divide the training data into a training set and a testing set 2415, These are all the categorical features in our data 23068, Command Center feature engineering pipeline classifier 35869, Training 43389, we choose one of these successful examples and plot its related perturbed image over a range of epsilons 33828, You can also select to drop the rows only if all of the values in the row are missing 15211, Several features requires imputation and normalization 33333, FEATURE 6 of MINIMUM PAYMENTS MISSED 38428, Improving the CNN architecture vol 2 3249, Distribution of the Target variable Sales Price 7984, Merge Porchs 11378, Lasso 37185, Importing important libraries 8753, Estimator 16878, Age VS Survival 33879, TotalBsmtSF Total square feet of basement area 19389, Final model to train and predict 37428, Distribution of no punctuations in tweets font 38152, we get to define a space of multiple configurations 1402, Embarked vs Survived 14259, Converting String values numeric 30366, Test cv2 without conversion 19606, 7 columns have only one value 0 and can be removed 41421, Deploy models on validation set choose the best one 1558, Women and children first goes the famous saying 10531, Outliers 8101, Feature Engineering 3028, We use the kfolds Cross Validation function where K 10 which is the number of holdout sets 17963, Dedicated datatypes for strings 9399, Once finished or even if you interrupt the process in the middle you find the best parameters stored in your Trial object 1673, use our logistic regression model to make predictions Remember we are still in our local validation scheme e we do predict Survival but without touching the testing dataset yet 20245, Extract the titles from the names 19527, Performing Reduce 21137, We want to apply hot encoding so we needed to check the number of levels in our data set 2549, Age creating bins as we know young were last to be rescued so lets explore this relation 6988, Number of fireplaces 30630, Correlation analysis 32031, Do the same prediction and replacement also in test X 1427, SVM 36897, Adadelta 31762, Prediction 36758, we can randomly visualize some images from the training set by plottimg them 43133, Fatalities 3 Best Predicted 8121, AdaBoost 2697, start by imputing features with less than five missing values 39765, Here I also use a CountVectorizer and a TruncatedSVD with 9 components to identify the nine main topics of insincere questions but with the parameter ngram range set at for the CountVectorizer 38630, use 10 of the train data as validation for early stopping 40443, MSSubClass 23316, Fix category 22139, TotalBsmtSF Total square feet of basement area 11642, Artificial Neural Network 28926, Sample 23290, All 41212, We are pretty close to the final prediction If we now apply Naive Bayes we can get the final prediction by multiplying these 201 terms as following 39263, Import data 40652, use validation fold models to predict on test set 23819, separate dataframe to study only categorical features and there mutual relationship and also the one with target column y 3007, strong Isolation Forest strong font div 32682, MultiClass Classification 31733, One Hot Encoding 20694, How to Use Advanced Model Features 39403, Binary Features Exploration 5415, Without a doubt the proportion of women survived are much higher than that of men 15870, Cross validation 31513, Class distribution 35553, Bagging 6679, Gradient Boosting 43269, Instanciando o objeto rf a partir da classe RandomForestRegressor 8930, Masonry veneer Features 28508, This is the code for GUI application that i have made for predicting numbers 16239, The fourth classification method is DecisionTreeClassifier 645, The 1 means that our boundaries are slightly shifted in terms of the real Fare 26090, Building new model and using batch normalization 9847, Continuous variables 41239, Ridge and SVM classifier for text data 38462, create a method to process any input string 15994, Searching the best params for Logistic Regression 29886, Showing Confusion Matrices 12171, Feature Engineering 8851, Submission 13870, We split the dataframe to get our original passenger data from the training set and the test set 29023, How about a nice boxplot for our numerics 1767, Do the same for test data 7534, Most of the people paid 0 80 Fare Fare varies based on Pclass and Survival Survived people paid higher fare than people who died we need to utilise fare column Since Fare as an integer column not be usefull Lets make it Categorical 13214, Evaluating our model with ROC curve and AUC 12528, Before doing it dig deeply into the data description 42820, Model 4690, Ordered 5256, Forecasting Model Experiments 36791, Secondly we try the Porter Stemmer 27342, Predicting labels and saving in csv file 19143, Model 3 with AdamOptimizer 40440, Training The Model 28512, Look at correations with SalePrice 8437, Kitchen Quality Miss Values Treatment 18634, the first holder date starts from January 1995 28815, Loading our Model 7335, Encoding sex feamle 0 and male 1 2362, Probability Predictions 15751, Clean Data 23315, Load data 10619, Looking at outliers in Class 1 it is obvious that mean of Fare is highly affected with these values 17704, DROP THE UNNECESSARY COLUMNS 32029, We predict age data for all rows in train X We feed all columns except Age as input 1076, Pre processing 35185, Correlation Analysis 23534, Please upvote if you like 36618, Using Sci kit Learn library 9896, give a trashold value for family size 8044, Cabin 25046, Aisle Reorder ratio 2469, Feature Importance 24111, Creating Model 13830, Title 5167, FE based on the my kernel Titanic Comparison popular models comparison popular models 40952, Correlation Heatmap of the Second Level Training set 3031, we use 10 fold stacking we first split the training data into 10 folds 12180, There is high percentage of values missing in the age feature so it makes no sense to fill these gaps with the same value as we did before 6025, We have skewed target so we need to transofmation 23085, Train your model with your trained dataset 27777, Using External Data From Keras Datasets to use that data as training data 20453, Go to top font 7404, take a look at our response variable SalePrice 25507, IMPORTING MODULES 11763, Stacking Models 20741, GarageType column 12817, Out of 342 survived travellers there are 233 female and 109 male 9120, If we took the log of the LotArea would this make its distribution more normal 15290, Creating categories based on Fare of passangers 17640, As some algorithms such as KNN SVM are sensitive to the scaling of the data here we also apply standard scaling to the data 38662, Passenger Class 34641, Multinomial NaiveBayes Classifier 18343, There is an order present in variable values Excellent Average Typical Fair Poor these were incoded to hold an order with the best group having highest number 15095, Stacked Model Learning Curve 35478, Calculating Area and Parameter of cancerous part of cell 30902, deal with regionidzip first 2638, This tells 3 value occurs 491 times 1 value occurs 207 times etc 14448, go to top of section engr2 26182, Feature Engineering 2751, Another way to fill categorical values is to use ffill or bfill 5544, Create Hyper tuned model 4635, Groupby and aggregation 3531, With qualitative variables we can check distribution of SalePrice with respect to variable values and enumerate them 2297, When is it an array When is it a Dataframe note the difference vs 22017, The most important feature for XGBoost is var According to a Kaggle form post customer satisfaction forums t data dictionary post 18717, After eyeballing the graph let s choose a maximum learning rate 17889, Mapping rare titles to rare 3456, To tune the parameters of the tree model we ll use a grid search 6449, Hyper Tunnning 19046, We can now explore the distribution of the data 11385, Engine s status analysis 13528, Manual FE 22691, Train and predict 30465, Confusion Matrix 36851, print Classification Report and Accuracy 22490, Bonus6 Chord diagram in Python 35751, Some ML algorithms need scaled data such as those based on gradient boosting 12373, Separating categorical and continuous data fields 27928, Before going on to form the data pipeline I ll have a look at some of the images in the dataset and visualise them with their labels 29701, lets take them through one of the kernels in first maxpooling layer 5590, Factorized 2 of the column whic are Sex and Embarked 42454, Handling imbalanced classes 8563, Finding Unique Values 2492, Observations 6979, Special cases 23408, Keras Model Starts Here 42656, In order to understand better the predictive power of single features we compare the univariate distributions of the most important features 13065, Before we try various models the data needs some additional pre processing Specifically we should covert categorical features to numerical values Sex Embarked 4553, BsmtQual BsmtCond BsmtExposure BsmtFinType1 and BsmtFinType2 For all these categorical basement related features NaN means that there is no basement 19530, Creating RDD from File 14637, work on filling the missing values for Age 10949, First five variables 17941, Name 5861, TODO Intro for K NN model with PCA 3165, Almost done with the data preparation but we need to express SalePrice as a logarithm since that s what the official error metric uses 27111, We have more missing values in test dataset than train dataset 6109, We fill missing LotFrontage values with square root of LotArea 7000, Unfinished square feet of basement area 32327, log error 29603, Examining the Model 14495, going for KNN 5930, Missing value function 36242, First sale There are multiple items sale at first time which shfit features are not covered The mean features group by category type subtype shop and city are created 19009, Plot feature importance 7233, Ridge Regression 9178, It is confusing to me that the minimum of this is 2 5057, On average the properties were 37 years old at the time of sale 11703, Scenario 1 Weighted Samples 15418, let s have a look if fares are a good predictor of survival rates 37769, To demonstrate the capabilities of this feature let us define method which evaluates pi using random generated data points and then look for ways to optimize 14711, Fitting here is a bit more involved then usual 32222, Looks like a lot of zeros 15628, Embarked categories 42113, That s pretty good just two classes but the positive class makes just over 6 of the total 32561, Diff Source 14285, Model evaluation with tuned parameters using cross validation 15321, Lets work out with the Cabin numbers 16044, Getting the data 23018, Event Pattern Analysis 18700, we ll use fastai s ImageCleaner Jupyter widget to re label delete images which are mislabeled noisy irrelevant 37156, Below I have extracted out the top 8 layer outputs of the model 40694, now the demand is highest for bins 6 and 7 which is about tempearure 30 35 bin 6 and 35 40 bin 7 37985, However trying too many combinations might explode kaggle kernel 14808, Find Missing Value 40859, Creating New Features 20220, Plot the validation loss and training loss 21511, Playing with brightness and constrast 35771, xgb reference 18533, Investigate the target variable SalePrice 37654, Visualize accuracies and losses 30698, Treinar um Modelo de Autoencoder 6260, Both males and females have decreasing chances of survival the lower class they were in 5953, Transforming Sex 35208, We started checking with a forest model made of 300 trees 22340, Corpus 8979, Given our new features we can drop Exterior1st and Exterior2nd 40623, Add a list of columns to be dropped and id columns to be used by our data processing pipeline 15716, It looks similar to the train dataset and includes all same columns except Survived which we need to predict 39875, Neighborhood 20817, Feature Space 32249, WELL WELL WELL 16334, Logistic Regression 7947, Check the model loss score here evolution RMSE 11391, Looks like first class passengers are older than the second and first class passengers 8949, Fixing Functional 35739, Do some PCA for the dataset to remove some of the collinearity 36699, Defining the model 7496, Use different models to predict y 22003, Run the next code cell to get the MAE for this approach 27981, Imbalanced dataset is relevant primarily in the context of supervised machine learning involving two or more classes 1251, The SalePrice is now normally distributed excellent 38478, Convolution 4660, The target variable is right skewed As linear models love normally distributed data 31048, Length of characters 43344, from here our Neural Network part starts 7451, Distribution of survival rate class wise 1204, Gridsearching gave me optimal parameters for XGBoost 13299, It s solution generate tuned DecisionTreeClassifier by the GridSearchCV from kernels 20905, Plot images 25285, we can figure out what ideal learning rates are 9326, Let a model help you 38228, The second diagram presents the average number of incidents per hour for five of the crimes categories 36724, Define train function 660, Perceptron 9011, There are 3 rows in the testing set which have null PoolQC but contain a pool 33083, check if it worked 19314, Evaluation prediction and analysis 1122, According to the Kaggle data dictionary both SibSp and Parch relate to traveling with family 522, Seaborn Distplots 37751, that we got the random sample rows let us fetch them from the csv file 32646, Along with traditional libraries imported for tensor manipulation mathematical operations and graphics development specific machine learning modules are used in this exercise regressors ElasticNet LassoCV RidgeCV GradientBoostingRegressor SVR XGBoost StackingCVRegressor cross validation engines a scaler and metrics methods Comments on the choice of regressors are provided in Section p 14585, Embarked Missing info 7989, Work With Labels 9373, count the house which have no basement in train test 34541, Below is the plot of occurences of n grams in the duplicate and non duplicate sentences 18112, Quick EDA 36125, Categorical features are now encoded and we concat categorical and numerical features and make final clean prepared dataset 29810, Displaying FastText WordVector of given word 36249, very simple term frequency and document frequency extractor 13478, Our Logistic Regression is effective and easy to interpret but there are other ML techniques which could provide a more accurate prediction 17447, Check the Values 40465, Garage 38053, We can find the group of a var using the following functions 113, calculated fare 34697, Linear extrapolation to the next month 13953, Fill Missing Value 33092, Lasso regressor 19253, Overall daily sales 10677, most of the features are correlated with each other like Garage Cars and Garage Area 43307, Split training data into train and validation sets 32985, Decision tree regression 6561, Survival rate vs Siblings or Spouse on Board 39343, Co variance Matrix of standarized data 37207, Final Training and Prediction 23605, finally for the similarity 37371, K Nearest Neighbor 42427, Max Floor Vs Price Doc 8604, Null in test data 1220, Uniqe 19929, check for null and missing values 6287, we take the training dataset and target column that we have just built and we create samples of this 31231, Features with max value between 10 20 and min values between 10 20 18370, Transforming Demand to Normal Distribution 29320, Predicting using VGG16 34483, When using tensorflow as a backend with keras the keras CNN s require a 4D array or tensor with the following shape 21464, The earlier layers have more general purpose features 42987, Word Clouds generated42995 from non duplicate pair question s text 132, Using Cross validation 11117, Fit and Evaluate Random Forest Model 31901, To verify that the data is in the correct format and that you re ready to build and train the network let s display the first 25 images from the training set and display the class name below each image 40444, LotFrontage 22154, Besides that with a simple linear transformation we could supperpose two distributions making them more readable to our human eyes Take a look at var and var before and after we transform var 39842, Visualizing the sunspot in 2019 24575, For Google Colab 22712, Training different PCA Models 17015, Ticket 981, whole thing 11183, Compare Train to Test data verify the distributions look similar maybe add probablity plots per feature with train and test on same chart 17593, we ll merge our data 7022, Quality of second finished area 5476, Check to make sure the addition of the components equal the prediction below 18598, Another trick we ve used here is adding the cycle mult parameter 2394, KNNImputer 37814, Target Distribution 30940, Visuallizing Interest Level Vs Bathroom 24392, Modeling 7426, Identify the most important features with names on the first 25 principle components 38061, We have to clean some special words inside the cleaning text process 11076, Out of fold predictions 34796, Lasso Regression 15971, Parch Feature 12314, Outlier removal 34635, e are going to convert cateforical data to categorical columns 28246, you can also check missing values like this without need of function 1002, This looks much better Now we convert the categories to numeric ones to feed them into our model 18819, This model needs improvement run cross validate on it 6006, Data Numeric 12643, Handling Categorical Text Data 17009, Maximum age is 80 and minimum age is 0 38304, To handle null values we take a random value between minimum and maximum value of each column and use that random value for imputation 2991, Missing Data Assessment 1547, Creating Submission File 30586, Counts of Bureau Dataframe 27904, MSSubClass LotFrontAge OverallQual OverallConditions are looking like a categorical value 12986, Simple Logistic Regression Model 7501, Proceed with rest of data using jointplots 40544, LinearDiscriminantAnalysis 20684, Functional Model API Advanced 41869, XGBoost using Scikit Learn API 10936, Numerical distributions before Data imputaion 6403, Linear Regression 37546, KERAS MODEL 17756, Training Final Classifier 7499, OverallQual and Fireplaces look fine since the better the quality and number of fireplaces the more expensive the house 3260, strong Identifying the total percentage of the missing values in both the data set exculding the target variable p strong 36888, dense 3 669, Ada Boost 27355, Starting with a base model 11474, Comparing Models 4292, Create an intercept 575, submission scores 22627, Preparing Test Data for Prediction 26822, check now the distribution of the mean values per columns in the train and test datasets 41734, Load the pretrained embeddings 6175, The order of passengers are highest for 3rd class then 1st class and then came 2nd class at the lowest 7556, Gradient Boost 18545, The training part contains information about 891 passengers described by 12 variables including 1 target variable 21608, Accesing the groups of a groupby object get group 36766, Data augmentation is a really powerful technique 42809, Features importance 21426, BsmtFinSF1 29046, Subtracting the median blur image from the original 8821, for these rare title we ll convert them to Others except Mme be converted to Mrs Ms and Mlle to Miss 39884, Prediction with Linear Model Ridge 8392, Seting the type of some categorical features 19421, Few more things to notice 4271, Electrical 25007, Adding Target Lags 32786, let s run the frequency encoding function 24915, Well we have to appreaciate India in maintaining constant cases 22875, Prediciting the Outputs 7470, Loading the csv files 12777, Machine Learing 28664, LotConfig 1918, All missing value indicate that particular house doesn t have an alley access 7543, lets convert categorical values into dummy variable and Scaling 4969, Title 40262, 1st Floor Surface Area 28151, Well this is not exactly what we were hoping for 30608, Test One 5032, Import libraries and set globals 38250, Correlation Analysis 6392, Columns 33681, Difference Hours font 39104, Applying Latent Dirichlet Allocation LDA models 163, Our second step is to check for missing values 30939, Visualizing Interest Level Vs Price 7794, not log the data since a neural network is quite good at working with non linear data 3817, sample mean population mean standard error 18577, Fare 6231, Final feature selection 14362, Most of the Passengers came without any sibling or spouse 6078, Age 25813, Output 734, I chose a p value of less than 0 37293, Exploring Target Column 22144, Data Preprocessing 3850, Feature Fare 22289, Identify categorical variables 14727, Normalize the Data 34858, Prediction 18831, XGBoost 14836, Pclass 1136, Model evaluation based on simple train test split using train test split function 40404, Training using Stratified GroupKFold 15144, We can check the correlation between Ticket number length and Sex 2108, Feature engineering and Feature selection 15622, Deck feature 405, XGB 13864, we extract the numeric portion of the ticket We create extra features such as 2733, Loading the libraries and data 24338, Average base models according to their weights 4889, Did you recognize something yes We can get the alphabets first letter by running regular expression 33348, Question 1 Create a plot with the moving average of total sales 7 days and the variation on the second axis 21399, Distribution of target variable 12841, Random Forest 15484, Test Against Cross Validation Set and Test Set and Analyse Performance Metrics F1 Score 10579, Evaluating ROC metric 7907, th model Gradient Boosting Classifier 9697, Observations 36012, Vocab 34654, Renaming and merging some of the types 8985, This may be a useful feature so I am going to keep it 6223, for categorical columns filling NaN with most frequent value of the column 5374, Embedded Methods 8359, FEATURE SELECTION AND ENGINEERING 5228, ICA 7603, sklearn pipeline Pipeline 4039, now look at the SalePrice variation on different categories of categorical variables columns 31003, Lets use XGBoost to assess importance 26721, For CA 17848, n estimators number of trees in the foreset 119, age group 22521, Split the full sample into train test 80 20 26557, Compile the Model 7781, value counts with default parameters 4737, Missing value of each columns 15706, Fare vs Survived 40287, Prediction 35670, Ordinal features 38743, Exploratory Data Analysis 14184, Treating missing values 5248, Setting Model Data and Log Transforming the Target 21256, Get Ratings By Sale Count 22324, Convert to Lower Case 3679, Sales price corr with new features 16433, DecisionTreeRegressor 2495, The chances for survival for Port C is highest around 0 5034, We have 81 columns SalePrice is the target variable that we want to predict Id is just an index that we can ignore we have 79 features to predict from 7837, You already know and use mean imputing mode imputing 12517, Checking the NaN values 14078, Pclass 37337, Save model weights and load model weights 22903, Most frequent words and bigrams 24514, We can now create a callback which can be passed to the Learner 26183, In this notebook I am just going to scale the data not making any new columns 14368, Analyzing Feature Fare 22841, Viola That worked 8452, Identify and treat multicollinearity 31227, var 12 var 15 var 25 var 34 var 43 var 108 var 125 have very low range of values further elaborated by the histogram below 8570, Grouping Rows by Values 19600, Last Preparation 2355, Support Vector Classifier 38111, The last 28 days are used for validation 33861, Machine Learning models 798, use our Sex and Title features as an example and calculate how much each split decrease the overall weighted Gini Impurity 41481, Drop the following columns 7912, We re gonna remove the principal outliers in the scatter plots of GrLivRea GarageArea TotalBsmtSF 1stFlrSF MasVnrArea TotRmsAbvGrd vs SalePrice TODO 9002, I want to create a columns that tells what type of Tier neigborhood the house is located in 18983, Display more than one plot of different types and arrange by row and columns 12264, Scikit learn implementation 16584, Fill missing Age 13382, Outlier Detection 36434, See which directories have you got 20926, reshaping into image shape images vertical height horizontal width colors 26659, Try to do the same model in Scikit learn 15975, With this feature happens the same that with Age feature 18518, Reshape 34850, Feature Importance 17921, The columns SibSp Parch Sex Embarked and Pclass contain categorical data Machine Learning Algorithms cannot process categorical data So one hot encoding is applied to these columns in order to convert the data into numbers One hot encoding is done below using pandas get dummies it cannot process 16924, eXtreme Gradient Boosting XGBoost 3595, Lasso 34687, Adding all the zero values 42953, Fare Feature 5395, Here I put training set and testing set together so that I can do preprocess at the same time and after data imputation I copy a set of training set so that I can do EDA with it 42347, We first have to transform the dataset into the ideal form in order to make XGboost running 14467, go to top of document top 453, Cross Validation 21664, Add a prefix or suffix to all columns 32012, Convert Categorical Features 29682, Data normalization in case of CNN and images helps because it makes convergence faster 30765, Score feature removal for different thresholds 14858, Since from the EDA I remember that we have missing values in both train and test data and multiple categorical variables to deal with I decided to use pipelines to simplify all the work 14186, Feature Engineering 17842, We check the features importance 11294, Label encoding for categorical variables with ordered values ordinal 4745, Numerical Features 43132, Confirmed Cases 3 Worst Predicted 22868, Pooling Layer 32503, Model 2 Using Transfer Learning Extracted VGG 19 features 16150, filling missing values 561, submission for random forest 42760, Fare Distribution 12279, For the evaluation metric of feature importance I used MSE of pertutaed data MSE of original data MSE of original data 33101, Stacking models 39883, Importance 34281, view distribution of continuous features using boxplot 18311, Mean Encoding for shop id 4429, Features that have too low variance can negatively impact the model so we need to remove them by the number of repetitive equal values 22499, Lets start import tensorflow 40662, in just base implementations XGBoost is the winner 19293, Data Generator 33838, Matrix 14670, KNN 10541, Merging numerical and categorical data 28064, The basic idea for age imputation is to take the title of the people from the name column and impute with the average age of the group of people with that title 3263, strong strong strong Visualising the missing values along with there respective coulmns p strong 8515, Creating Dummy Variables 9623, Model Building 40391, Defining our Learning Rate Scheduler function 38963, In this section we do the actual training comining the previous defined functions to a full pipeline 12989, Prediction 9739, Throughout this notebook I pretend that testing set is never exists until the model is trained to simulates the real life scenario where the data to be predicted comes later 26684, convert categorical using LabelEncoder 26726, Plotting monthly sales time series across the 3 states 1189, Creating Dummies 643, remind ourselves of the distribution of Fare with respect to Pclass 29700, lets take them through one of the kernels in next convolutional layer 12239, If statements 10902, Grid Search for SVM 30392, Targets 23486, TFIDF can be generated at word character or even N gram level 13089, Since PassengerId Name and Ticket columns do not provide any relevant information in predicting the survival of a passenger we can delete the columns 3016, These visualisation helps determine which values need to be imputed We impute them by proceeding sequentially through features with missing values 27922, Find the tree size where mean absolute error is minimum 26800, Confusion matrix 4978, Exploring Tickets 13509, Curve 19297, Data statistics 7377, To deal with the duplicates I define the function dupl drop that removes duplicates both in the columns PassengerId and WikiId 2115, Since we already have GarageCars to describe the Garage and this new feature is very correlated with the basement SF we could consider if it is better to use it and drop the original two 43037, CNN Model 8998, I am just interested in the values of the features that have 90 of the same value 4506, Magnifying Further 26574, Setting up the data pipeline 24414, Housing Internal Characteristics 2916, I print a heatmap with the correlation coefficients between the features 40033, take a look at some example images and their augmented couterparts in train 41994, iloc To read a single row 36943, Title IsMarried 26078, Split data 12383, Removing outliers from continuous data fields 20490, Contract status 33296, Ticket Cleaner 18084, Plot the darkest images 10352, We use lasso regression 8841, Prepare categorical 41414, Basic modelling LGB 2735, Taking a first look at our data gives us a rough idea about the variables and the kind of data it holds 15411, leave the missing Cabin values unfilled and let s separate out the training and test data again 18247, Identify the missing values 13745, XGBoost 37002, check the total number of unique customers in the three datasets 16502, Linear Support Vector Machine 26639, Test data prediction 32540, Extra Analysis 9205, Gender Distribution 36529, I looked at every pair of images here and only the last pair was a pair of images different in an important way 36466, Images from ARVALIS Plant Institute 2 19256, Sub sample train set to get only the last year of data and reduce training time 36448, Modeling 14787, Embarked 26955, Check missing value from data 6702, EDA is a way of Visualizing Summarizing and interpreting the information that is hidden in rows and column format 10615, Embarked column in training data 11679, There are two numerical variables that have missing values namely Age and Fare columns 13050, Cross Validation Strategy 28649, Fence 25386, preprocess of data 11126, Prepare Submission file 22442, Pairplot 35828, Feature Engnieering 16969, Outliers detection 2922, K NearestNeighbors 804, Settings and switches 43322, Normalizing Images 11255, The train data now be split into two parts which we call training and validation 35105, Creating multilabels 39965, Categorical Encoding 34482, Original version takes Gb of memory 35886, Ordinal encoding of remaining categoricals 17947, Encode Age 24935, Feature transformations 650, We might even be at a stage now where we can investigate the few outliers more in detail 4255, There are 19 columns with nuls in the training set 10247, Go to Contents Menu 19246, Albumentations 35207, Random Forest 19956, The first letter of the cabin indicates the Desk i choosed to keep this information only since it indicates the probable location of the passenger in the Titanic 18605, tfms stands for transformations tfms from model takes care of resizing image cropping initial normalization creating data with mean stdev of 0 1 and more 15230, LGBM Model 42519, Reshape To Match The Keras s Expectations 22002, Drop columns with categorical data 15344, GRADIENT BOOSTING 22718, Creating the labels matrix for the plot 12770, It is time to make submission 20295, Correlation Between The Features 12604, use chi square test to understand relationship between categorical variables and target variable 1257, Feature transformations 13480, try another method a decision tree 14556, Heatmap font 14482, Correlation Analysis 16372, Exploring Fare 38567, Dimensionality Reduction Techniques 31667, Evaluation of model with the best classifier 41774, Show network 35467, Visualiza the skin cancer seborrheic keratosis 22090, Build Train and Test Data loaders 27249, Obtain the training set 6030, Check Skewness and fit transormations if needed 13901, Graph on passenger survival Pclass wise 30273, hr 21440, Overall Rating 36494, Save folds 23209, Findings Boosting method can t beat best boosting base learner gbc Though it could beat if we would have optimized xgbc If you have time and infrastructure you can tune xgbc s hyperparameters compare boosting accuracy with its base models accuracy 3923, Decision Tree Classifier 3713, LonFrontage 17256, Looking into Training and Testing Data 33098, XGBoost regressor 3887, Mean 8929, Basement Features 38175, Pycaret needs an input to be entered after the next command and since kaggle doesnt support commenting it out 6919, Learning Validation Curves Jtrain vs Jcv to help identify under or overfitting 4788, Handling missing values 12432, Using regex 24321, I want to do a long list of value mapping 34664, Around 90 of all daily sales include only one item 19712, We have equal number of cats and dog photos to train from 38635, the pooling layer 7959, Submission 6833, Feature Interactions 6444, Numerical variable which are actually categorical 16234, we perform transform function which is known as transformer A Transformer is an abstraction that includes feature transformers and learned models Technically a Transformer implements a method transform which converts one DataFrame into another generally by appending one or more columns 21128, Now we look at categorical variables For this the idea be very similar so starting from NaN s 20758, SaleCondtion column 28362, Importing Libraries 29193, Extreme Gradient Boosting 36417, Correlation Matrix 30708, now do the same with COCO bounding box format 33020, As we know the images come in a grayscale format where all the values are between a good thing you should do is standarize the data which makes it easier for the model to converge 19185, We know which features are the most significant for our model so we check the distribution of those features with respect to the target variable in bar plot scatter plot with linear fit and finally box plots to know the statistics 6169, Embarked 41830, Visualization of learning process for single model 35468, Visualiza the skin cancer lentigo NOS 4869, Label Encoding of some categorical features 38851, Feature Engineering 9900, Pclass 17660, Observations 8622, GarageCars vs SalePrice 24657, I use only new cases and new death s dynamics to make the prediction 30597, At this point we save both the training and testing data 30997, Feature primitives 5025, Ordinary Least Squares OLS Linear Regression No Regularization 14164, we can work out some meaningful interpretation from the Cabin column 27179, Similarly we can tune other parameters for better accuracy 37502, the correlation in between year sold and sold price is not that much variate 26409, Most passengers have neither children nor parents on board 28332, Types Of Features 4851, XGBoost 751, Before training let s perform min max scaling 37802, Regression Evaluation Metrics 35592, Freezing Layers 13882, This is the information we have in the training data set 17915, SCATTERPLOT ANALYSIS OF PASSENGERS AGES IN EACH CLASS 16576, If you are a begginer you can leave this portion of creating FamilySurv and come back later when start unserstanding 34540, n grams are the continuous sequence of words They can be single words if n is equal to 1 or continuous sequence of two words if n is equal to 2 and so on 36375, In percentage 24542, Again let s exclude the dominant product 16395, docs version indexing htmlboolean indexing 39344, Findind top 2 eigen value and corresponding eigen vectors for projection in 2 D 36116, Lets fill the missing values 19156, Feature preselection 38038, Light Gradient Boosting Method 28268, NorthRidge Northridge Heights comprise of moderate to hmost expensive houses 24805, explore data 34109, Hospital beds for covid patients across state 14899, Visualize Age Fare Family size and Survived 13688, Embarked I decided to drop the passengers who didn t embark since modeling based on their data would act like noise in my opinion I feel that they can t reliably tell us about the survived not survived ouptut 34288, normalize data with min max 10870, 3D visualization 35485, CutMix data augmentation 32357, Tweaking threshold 10263, Restore training and test sets 6807, Logistic regression is the hello world of machine learning algorithms It is very simple to understand how it works here logistic regression 9b02c2aec102 is a good article which cover theory of this algorithm 31237, Before concluding let s do a check of whether feature values in test and train comes from the same sampling distribution 17591, We use heatmap for plotting the missimg values 25576, GrLiveArea 16961, Feature engineering 32331, Univariate Analysis 32593, We can also find the best scores for plotting the best hyperparameter values 58, Making the dataset ready for the model 18494, Developing The Model Define a Performance Metric 25802, isolate the entries relative to these 100 managers with the interest level column as well 21636, Access numpy within pandas without importing numpy as np 26472, Logistic Regression 14250, The chances for survival for Port C is highest around 0 4653, Frequency Distribution 25994, Creating a SparkSession 19391, Extract the ordered products in each order 31264, Sample sales snippets 11629, Evaluate the model 39208, Research Question Is there a significant difference between the length of text of real and not real tweet 38275, train our model with our padded data pad docs and label with epochs and batch size 27943, Extract features and make predictions on the test data 28940, We can even infer that the passengers who embarked at Southampton had a lower chance of survival 14609, Linear Discrimination 36728, Actually I noticed the discussion about 35063, Complexity graph of Solution 3 3681, Explore the Target Variable 38844, Most of the house built area is less than 20K Sq Ft 38055, Using GridSearchCV to find the best parameters for SVM 12387, One hot encoding of all purely categorical data columns 21317, Target varibale l logerror 4022, Garage Columns 20835, We ll be applying this to a subset of columns 32920, Wavenet Model 31793, First let us define the DenseNet backbone 41357, BsmtFinType2 is not useful for us for correlation relation between price 5893, Some Missing values in test data 26507, we add a softmax layer the same one if we use just a simple 15977, We inpute the missing value with the mode 21804, SHAP values 11908, People with destination C Cherbourg survived with highest percentage 35324, Trying to use ensamble method the simplest bagging 23397, Test Images 13272, Statement The boys from the small families Family of the third class cabins who were sitting in Southampton all survived 1903, Submission 39317, XGBRegressor training for predictions 21570, Save memory by fixing your date 10500, Sex 39720, Playing with the trained Word2Vec model 13904, Lets try XGB Classifier to train and predict the data 23559, There are also records with floor values greater than the maximum floors in the building 6289, Ensemble 11042, Score the models with crossvalidation 32126, How to find the count of unique values in a numpy array 15860, The following cell uses Baian hyperparameter optimization to search for the best parameters for a Random Forest model 11670, It looks like females are more most likely to survive than male 13093, Cramer s V is more appropriate than Pearson correlation to find correlation between two nominal variables 28329, identifying the categerical and numerical Variable 7814, NA value 1152, Model EvaluatioF 15585, On the leaderboard this scores 0 29091, Comparison to the Original weights 23932, We can then replace the pre trained head 16134, Final Predictions for Competition 95, Combined Feature Relations 41267, Another way is to use gridspec 23173, Encoding Categorical Variables 25794, We can create Band for Age and Fare let s create it 22351, XGBoost 20833, Same process for Promo dates 4018, MasVnrType and MasVnrArea Imputation 12266, Download datasets 29220, Thus we are going to use 9 Principle components to preserve 96 31933, visualize some of the flase predictions to try to get more of an understanding of the model s misclassifications 9431, Boxen Plot 37436, Wordcloud for selected text font 10454, LightGBM 23695, Fine tunning 11011, We are going to create a feature called Salutation depending on the title the passengers had in their name 38490, Errors font 33200, Defining a Small Convnet 1877, Nice No more missing values 21450, Categorical Numeric 11348, Logistic Regression 25169, Simple fuction to perform stemming while removing StopWords 11637, Random forrest 23508, There are 17 elements in the equivalence class 14111, center CatPlot center 9965, Distribution of Sale Price 21251, Model Building 9808, Missingno library offers a very nice way to visualize the distribution of NaN values Missingno is a Python library and compatible with Pandas 8009, Train Random Forest Regression 15693, Gender vs Survived 43020, RandomForest 33834, MEDIAN Suitable for continuous data with outliers 21200, Merge all functions into a model 17530, Create Age bads 23236, Find the best parameter value 5677, Parch and SibSp 33486, Linear Regression for one country 30337, We create the actual validation object 1520, We base this part of the exploration on the 25950, Feature Selection 1588, We can also make feature transformation For example we could transform the Age feature in order to simplify it We could distinguish the youngsters age less than 18 years from the adults 15055, Prepare Training 8145, Feature Engineering 8109, Fare 7715, Outliers affect the mean value of the data but have little effect on the median or mode of a given set of data 26991, Starting point 41188, let s print missing value counts for numerical and categorical columns for merged data 22835, Well then this means there are certain items that are present in the test set but completely absent in the training set Can we put a number to this to get an intuition 16137, import python lib for visualization 1951, We remove the columns which have more than 15 of missing data 23717, LotArea is highly skewed variable having skewness values 10 42619, Making Predictions 1102, Create datasets 7683, We do the same thing with SalePrice Target values column we localize those outliers and make sure they are the right outliers to remove 18342, DATATYPE CORRECTION 20682, Sequential Model API Simple 4259, Non null unique values 23631, get the validation sample predictions and also get the best threshold for F1 score 23446, Season 5205, now check again the dimentions of our training set after engineering the catagorical features using the get dummies function 11817, SalePrice and Overall Condition 32310, Relation between Survival and Passenger Class 5295, I can split the data into training set and test set 27126, Garage 18551, The mean age of survived passenger is 28 26114, We can drop PoolQC MiscFeature Alley and Fence features because they have more than 80 of missing values 19163, Brand name label features 38463, Custom testcases 29800, Fetch most similar words wrt any given word 1798, SalePrice vs 1stFlrSF 35164, Compile 10 times and get statistics 32083, Check for Missing Values 40098, Add missing income 33268, Create submission file 28726, Descriptions of the top 5k most exepensive sales 11120, Plot Number of Features vs Model Performance 41761, You can actually follow along as the search goes on 18502, Naive Submission 1380, to Finish our exploration I create a new column to with familiees size 27927, Cross Validation 4132, Label Encoding Categorical Data 23201, Looking at this plot one thing we can say that a linear decision boundary not be a good choice to separate these two classes we would train our models on this 2d transformed samples to visualize decision regions created by them 36777, Here format sentence changes a piece of text in this case a tweet into a dictionary of words mapped to True booleans 6977, concatenate train and test sets into one so we can analyze everything and fill NaN s based on all dataset 41593, priors Has multiple instances of the orders that is each product in an order is a separate row 42784, Training the model 1868, Refit with Important Interactions 2931, Plot Learning Curves of standart Algorithms 42880, Time plot 5833, Firstly lets convert string type values of object features to categorical values 24352, In order to make the optimizer converge faster and closest to the global minimum of the loss function i used an annealing method of the learning rate 6, focus on some of the categorical features with major count of missing values and replace them with None 42686, I want to draw two count bar plot for each object and int features 28501, Each image is 28 pixels in height and 28 pixels in width for a total of 784 pixels in total 10424, XGB 23418, we can predict the test values and save them 3766, Magic Weapon 1 Hyperparameter Tunning 3633, Age 17267, Data Visualization 38073, Stratificcation 32464, Models Data Prep 40303, Leaky Features Exploration 29732, let s the the key space we are going to optimize 6517, Categorial Variables 26690, AMT REQ CREDIT BUREAU TOTAL the total number of enquiries 5929, coordinates of point on scatter plot seaborn 43266, Importa fun o draw tree que serve para visualisarmos a rvore de decis o 20713, Alley column 41572, Making the functions to get the training and validation set from the Images 1531, Cabin Feature 12468, In general barplot is used for categorical variables while histogram density and boxplot are used for numerical data 7077, All the Title belongs to one kind of gender except for Dr 23474, Reading the test file 26978, Load libraries 23690, Take a look at some pictures is everything alright 41530, The digits in this dataset are more unifrom in those of the kannada dataset 14743, we have a 6 dimensional DataFrame and our KNN classifier should perform better 18338, This section i am going to zoom at variables which are corelated to target variable SalePrice 41242, Tensorflow model 19180, EDA on predictors 37568, Integer Columns Analysis 8099, Parch vs Survival 16732, concat datasets 5244, Dropping Unnecessary Features 10403, Confirm column overlap between train and test 9971, I created the area util feature my idea is Lot Area it s the house total area and others it s non raw area so i sum others area s and subtracted from the total area 8856, Calculate Shap values example 1 26418, The passengers have different titels which give information about their sex about their social status or about their age 22022, var36 is most of the times 99 or 33732, Splitting the data into train and validation set 6607, Decision Tree 11964, Prediction span 31857, Producing lags brings a lot of nulls 18519, Label encoding 807, shape info head and describe 36126, Different Models 17936, The more you board in Cherbourg the more likely you are to survive and the more likely you are to board in Southampton the more likely you are to die 40479, Gaussian Naive Bayes 42026, Split strings into numbers and words 36389, When using such encoding method there might be some category values in the test data that are missing from the train data 7954, Tuning on activation function 21235, Set class weights 31564, transform is None 11974, In the next step we deal with the numerical features 3721, PCA Principle component analysis 43402, The real code starts here 21569, Count the missing values 20092, Data Preparation 28668, Access 11620, Additional exploration 2511, Predictive Modeling 9646, Pretty much clustered in the range of 0 1000 GarageArea 20491, Payment type 14783, Ticket 12945, Age Column 3477, Model 2 ElasticNet 14509, Observations 10988, Build Our XGBoost Model 40070, Multi colinearity Numerical 29611, Plot random images with the labels 35927, Create submission file for Kaggle 12382, Box plot of all continuous data fields 40874, Retrain and Predict Using Best Hyperparameters 26086, The number of errors for the each digit 41459, The majority of females survived whereas the majority of males did not 34706, Cumulative shop revenue 39770, I was wondering wether the test dataset was an extract from the train one or completely different 31327, Process Test Data 12915, we are right that Age Cabin and Embarked contain missing values in training set 18313, assuming month of year plays a big role in number of items sold 21385, Fit the model 28263, Discrete features 30995, An EntitySet is a collection of entities and the relationships between them 43082, look to the top most correlated features besides the same feature pairs 5830, After dropping Id and other columns with High no 42687, Sometimes null data itself can be important feature 31897, Fit Model Using All Data 10658, Classification 36550, How do the scatter plots look like 31384, We be first importing the data and creating copies 32908, Model 12471, Outlier Treatment 2124, After this search the best configuration is given by 16106, Fare Feature 39164, We don t want that This would confuse our CNN 17357, Decision tree as a predictive model which maps features to conclusions about the target value 12487, Deployment 20561, simple cleaning the math tags 26933, This method help to obtaining a bag of means by vectorising the messages 12040, I m gonna put 0 in MiscVal for house which don t have any MiscFeature and None value for house with 0 in MiscValue and some value in MiscFeature 23873, Above are some of the most important features 19296, Find best threshold 4811, Exporting our submission dataset to a csv file 26819, Distribution for Standard Deviation pre 42143, Fashion MNIST 28864, The Encoder structure is very similar to an LSTM RNN network 5901, u Feature Importance u 36670, Text Pre processing 966, Second level learning model via XGBoost 5336, Diplay series with high low open and close points 39677, Initialize all the variables we created 34485, Create and Compile the model 36925, When I googled Stone Mrs George Nelson Martha Evelyn I found that she embarked from S Southampton with her maid Amelie Icard in this page Martha Evelyn Stone Titanic Survivor titanica org titanic survivor martha evelyn stone html 32942, Prepare the submission csv 37291, Reading Dataset 23027, Though the item is same sell price is slightly different in each store and each season 24676, Data Normalization 34835, Analysing the correlation between the feature and the sales price and select most correlated feature only 37818, Merge the train test to process everything atonce 34466, Expanding Window Statistics 27235, Calculate importance of each feature to iterate feature engineering 9848, I replace them by 0 23366, Training The Model 27471, Here we are just taking a list of text and combines them into one large chunk of text 257, Checking for number of clusters 36010, Datasets 29868, The images are actually quite big 42466, Checking the correlations between ordinal variables 28395, Null Value Management 8083, Setting Up Models 21639, Formatting different columns of a df using dictionaries 23956, Applying Random Forest Regressor Algorithm 30969, I think we re going to need a better approach Before we discuss alternatives let s walk through how we would actually use this grid and evaluate all the hyperparameters 40990, For example applying a lambda function 30410, LSTM models 8951, Fixing Garage 6859, Understanding Variables 12, Principal Component Analysis PCA 9953, Data Modeling 43038, Compile the Model 16388, Removing Less Important Features 12408, OverallQual 3898, Target Encoding 27293, Intervention by after days for SEIR model 15646, Bagging 31906, EVALUATING THE MODEL 26835, Hourly Rental Change 3011, strong Imputation strong font div 12448, it follows a linear correlation 25437, Split Training and Validation Set 21782, we would transform categorical value to their one hot encoding version 29117, check the the top 6 images with highest losses 9891, We can calculate the probability of people who could survive by looking at their genders 32870, Mean encoding 7110, we can evaluate all the models we just used 7026, Kitchen quality 33999, Right skew 11113, EDA Relation between each feature and saleprice 6885, Distribution of survived 1 is for survived and 0 is for not 35779, Averaged base models class 29974, Applying tokenization process 25750, Here is a function that crops a centered image to look like train size test size 0 23999, K Folds cross validator 24561, Distribution of products by activity index 38270, as we have applied all the cleaning steps so now its time to separate our data back 37096, Missing values 42359, Adding location to this sparse matrix 28426, In our test set we have 5100 sales in really new shop and no outdated shops but anyway it is good feature for future 13571, Ploting Family Size 18547, Check data for NA 29378, Transfer Learning 6270, These age bands look like suitable bins for natural categorisation 20321, the Python gods are really not happy with me for that hacky solution 23613, Work with missing values 14479, now we focus on which group from each Passenger class survived more and less 1172, Zoning is also interesting 31813, Get new image size and augment the image 150, Extra Trees Classifier 11016, We are going to guess missing values in Age feature using other features like Pclass and Gender 4379, Observing Graph plot there are such home exists with 3 fireplaces and there SalesPrice not much 6160, Remove the features which give us less then 0 7824, We have two models to fit the data 40142, Create submission output 15493, Hidden neurons 8704, LabelEncode the Categorical Features 14580, PassengerId Name and Ticket not play any role in Titanic survival chances 30932, How about longest questions 35359, Raw 18159, checking the number of row and columns present in the input file 20032, We would like to remove some of this redundancy 8839, Extract target variable 30340, These are the parameters that we are going to input to the previous function 14553, SibSp Siblings Spouses font 13205, let s separate our data for train and test 4888, corr Fare Pclass is the highest correlation in absolute numbers for Fare so we ll use Pclass again to impute the missing values 23804, Modelling with Generalized Additive Model GAM Whitebox Model 10511, create a New Attribute Alone which would be True if he she is travelling Alone 8112, MACHINE LEARNING 16469, Here I m not using HPT for LDA 19962, Hyperparameter tunning for best models 40253, Our Q Q Plot tells us that there is single outlier over 4000 which is causing major problems to the distribution 42399, Is there seasonality to the sales 16881, NewFeature FamilySize 13744, Comparison 18021, New dataframe Woman Child Group by Ticket 22153, WHAT I WOULD LIKE TO HAVE DONE IF I HAD MORE TIME 20288, Embarked 26783, How many combinations of GPS latitude and longitude 20678, Fit the Model 20927, Model 24262, Correlation analysis with histograms and pivot tables 41437, Pre processing tokenization 18692, After eyeballing the graph let s choose a maximum learning rate 26871, In the first layer we have 60 parameters 30368, Test SKImage 15312, Lets find out the percentage of Women and Men 25960, Top Selling Product Departments 23510, There are 2 elements in the class 29546, The main words are articles 5053, look at the time related features like building year year and month of sale 26073, Importing 27169, It is important to convert the categorical text data into model understandable numerical data 7214, Plotting some graphs for insights 38019, At its best what kind of predictions do the network trying to make 32812, Level 3 XGB 33857, Visualization using t SNE 36431, Model Training 8027, Let replace missing value with a variable X which means it s Unknown 5855, The two outliers we wish to remove are those with very low SalePrice relative to their GrLivArea square footage that are inconsistent with the trend of the rest of the data 38058, Checking target distribution 13854, Feature engineering 11819, We have Outliers in our data 22287, Take a look at your submission object now by calling 11752, we are going to take care of the skew that we noticed in some of our predictor variables earlier 35572, Lets look at a lot of different items 17878, Outlier Treatment 15391, Have a look at the test data 20357, This submission gets on the public leaderboard 22411, for gross income aka renta 8277, Importing Libraries for Modelling 31669, Preprocess Data 41527, The number 4 1372, cross our Pclass with the Age cat 20569, It helps in determining if higher class passengers had more survival rate than the lower class ones or vice versa 10415, Score pipeline 16742, modeling 28404, Neural Network 28305, Finding Root mean squred error for DecisionTreeRegressor 28684, MoSold 4520, RandomForestRegressor 8374, Ensemble 732, removing these two 24741, Outliers 38999, now shuffle the data set randomly We also have to make sure that x input in general sense and y output labels in general sense remain in sync during the shuffling 39246, Remove outliers in price data 2186, Validation curve 38778, Select investment sales from training set and generate frequency distribution 999, We also impute the Cabin missing values We use U for Undefined 18851, Visualize ROC on the Validation Set 4358, Feature 21372, Check missing data 6699, Lets create some swarm plots between Survived and Fare 21577, Aggregate you datetime by by and filter weekends 9427, stripplot 19592, city code 28199, we can finish up this part of speech tagging script by creating a function that run through and tag all of the parts of speech per sentence like so 14798, Random Forest Model 13305, Ensemble modeling 11214, Ensemble prediction 34254, Predict on Entire Data Set 33260, Data Preparation 40797, Checking date 26495, The corresponding labels are numbers between 0 and 9 describing which digit a given image is of 37416, PCA Example 4067, we can finally chain all of these transformations in a pipeline that be applied over our dataset before training testing and validating our models 37574, Important Variables 27552, Display interactive filter based on click over legend 24294, now we can plot that image 26069, Looking at the best probability achieved we have a digit that the model is 100 confident is a three 28937, About 65 of the passengers were male and 35 were female 40741, Large CNN 39224, Remove the duplicat columns from training set and test set 20293, Parch 12966, Embarked Sex Pclass and Survived 23792, Prediction and Submission 39848, Auto Correlation andal Correlation Graphs 30568, Aggregating Numeric Columns 40398, Fold 5 6456, PCA 10410, Run outlier detection on Overall Quality of 10 25389, Defining the model architecture Using ConVnets 36800, we create the chunk parser with the nltk RegexpParser class 17623, Fare per person 6678, Ada Boost Classifier 21365, Warmup Look at the Lonely Characters in selected text 8251, Univariate Analysis 41447, K Means Clustering 4771, 101 passengers from third class embarked from C 113 embarked from Q while 495 embarked from S 7523, Model 4 Exponential fit y A exp Bx 40323, add obtained user components as cluster features 11643, Cross validation 5906, Types of scoring in grid 17856, Create the ensamble framework 2516, K Nearest Neighbours KNN 4151, Random Sample imputation on Titanic dataset 20855, Some categorical variables have a lot more levels than others 15267, Split into Train and Test data 9769, Dummy Variables Encoding 4154, End of Distribution Imputation on Titanic dataset 11660, Decision Tree 41969, Convert Date Column data type from object to Date 16449, Cherbourg port is more save as compared to others 30143, Preparing Submission File 5302, The rest of the data processing is from the kernel by Boris Klyus 19003, Use test subset for early stopping criterion 197, LightGBM 22025, num var5 7915, Replace NaN values 9675, Tune leaves 2838, Libraries and Data 33603, Convert the data frames to numpy values 20267, Analysis of loss feature 10411, Remove the outliers from our data set 8989, the number of stories does not have a linear relationship with price 42727, draw the heatmap of float features 30635, Relationship between family members and survival rate 28375, Creating Training Testing Set 42575, The nested for loops below calculates every possible score way that our predicted values can produce and keeps track of which sum is built out of either a correct or wrong 13180, First let s take a look into Fare distribution 12962, Parch and Survived 21582, Correct the data types while importing the df 20073, Insights 42250, Perform feature selection and encoding of categorical columns 30590, Aggregated Stats of Bureau Balance by Client 15102, As expected those passengers with a title of Miss tend to be younger than those titled Mrs 12759, I explore the size of the dataset to compare it with the number of NANs in the dataset 32829, We are here at the data cleaning part 38248, find the top 5 RGB distributions in each of the 10 test images 33698, Yearly Series font 29870, Training 34844, Save Models 23267, Fare 30888, Final training and test predictions 16684, Analyze by pivoting features 42934, Saving the list of original features in a new list original features 32628, Naives Bayes Classifier 4465, We then assign social status titles to them for more in depth analysis 2357, Pipeline and Principal Components Analysis and Support Vector Classifier 32174, FLAT FEATURES EXAMPLE 26866, look at the effect of these two new filters 17726, pandas 39454, checking missing data in installments payments 42115, For our second model we ll use TF IDF with a logistic regression 39249, Import data 38651, Age 22005, If you now write code to 6307, Linear Support Vector Machine 97, This is another compelling facet grid illustrating four features relationship at once They are Embarked Age Survived Sex 17035, Rare values Pclass Sex Embarked 10963, KitchenQual 27054, Defining and fitting the Model 7228, LotFrontage 31536, MasVnrArea 38519, Pre processed selected text columns 5746, Checking survivors by sex 29060, Merging multiple outputs 18648, There are quite a few number of missing values present in this field 29759, Train the model 31283, Exponential smoothing 23601, Like I said I am compiling my understanding of gensim from a lot of sources and one of them used multiprocessing stating that it might be painfully slow doing otherwise 38090, Activation Function 36778, Training Data 2276, Building Machine Learning Models 8407, Garage areas and parking 40731, Try to compare the layer 1 output 28767, we create a function that plots the accuracys and losses of the model output 38988, Traing for positive sentiment 11663, Extra Trees 11033, Linear SVC 13747, Gradient Boosting 13899, Around 300 passengers survived and more than 500 died 7615, For some linear models QuantileTransformer gives best score for C 37023, Top 15 second levels categories with lowest prices 19725, For each item 30086, Naive Bayes Algorithm 21674, Memory reducer 8341, Moving on to neural network we simply use Keras for easy implementation of multi layer perceptron 4586, We also test the predictive power of the following features during model evaluation 6206, Logistic Regression 15734, ROC AUC Score 30282, Mortality Count 5 6799, Predi o modelo 41744, Some fancy pants code my teammate Alex made for the toxic comment challenge that I ve expanded on and adapted to this challenge 6019, Save Model 33324, Training and validation data generator 5604, Season 21106, Spilit training validation and test dataset 29373, Create Train Test 3645, Feature Engineering continued 23225, Here all the values which are True are the duplicate ones 250, Library and Data 11612, Embarked I fill missing value by most frequent appear value 8128, Correlation between values and SalePrice 4948, use the Box Cox transformation to deal with the highly skewed values 18377, R Square Error 3320, Version Added new plots from Seaborn release 37488, Multilayer Perceptron 21955, Confusion matrix 29540, we re gonna take 10 data from the training data and use it for data validation 967, Producing the Submission file 15265, Convert Features into Numerical Values 28492, Model fitting 16466, Pearson s 8842, For some categories the order is quite important like OverallQual 22295, Analysis and Submission 22474, Plotting with different scales using secondary Y axis 34847, Finding Duplicated columns 6661, Split the data into train and test set for classifcation predictions 41785, Optimization 10333, We conclude that 12525, so these are the columns which have missing value and have numeric data 16339, Random Forest 43137, Plot Loss Trend 41205, load our submission format and fill SalePrice columns with our predictions data 1289, Cross Validation 5 1 4874, Function for Scoring Training and Testing 37924, Final Prediction 37046, Stemming is the process of producing morphological variants of a root base word Stemming programs are commonly referred to as stemming algorithms or stemmers A stemming algorithm reduces the words chocolates chocolatey choco to the root word chocolate and retrieval retrieved retrieves reduce to the stem retrieve 14710, SVC Parameter Tuning with GridSearch 35770, XGBoost 35508, Outliers 6599, Visualize default tree optional 22944, This may not be the best feature for us to use because most of the data is considered null 35462, Visualiza the skin cancer at head neck 25809, Fit the Model with GPU 15033, Embarked with value of S have the most count of data set the null of Embarked to be S 37109, Code in python 21335, Chu n h a d li u 35879, Selling prices 39160, The last step is creating the submission file 43026, Model 3 Overfitting 12003, let s check the R Squared for lasso 2151, Imports 27565, ps ind 01 4639, Count of distinct categories in our variable but this time we don t want to count any nan values 15637, Re train the model on new features 36737, Comparing MAE values of all models 22642, Model 6 XGBoost 16108, Map Fare according to FareBand 13224, GridSearchCV for SVC 32945, Remove all special characters split by spaces and remove stopword entries in list for train and test 28140, Submitting the predicted labels as a csv file 20770, Tf idf space 5939, there are many columns that contains character values 23249, We do the same for test data as we do on train data 1386, Predicting X test 9023, There are 30717, One hot encoding 8409, Total Basement Area Vs 1st Flor Area 16668, Feature Transformation Categoric Variables 29963, Inference 1409, I need to replace a few titles with other values because these titles are not as popular and have a low frequency of occurrence in this dataset 6056, BsmtQual missing values for No Basement 27973, Mean Frequency 32713, Bidirectional LSTM 2663, Constant Features 9214, Family Survivor by Familymember 3778, Classifiers 38281, Imputing data in Basement related columns 15825, train and test data split 70 for training and 30 for testing 29467, Basic Feature Extraction 21239, visualize how our predictions look like 25752, Surprisingly it still can in some cases 20737, Electrical column 1625, Linear Regression with Ridge regularization L2 penalty 37978, Build the CNN model 13963, Pclass Survival probability 28327, identifying the missing value in previous application 5172, Each model is built using cross validation except LGBM The parameters of the model are selected to ensure the maximum matching of accuracy on the training and validation data A plot is being built for this purpose with learning curve learn org stable modules generated sklearn model selection learning curve html highlight learning curvesklearn model selection learning curve from sklearn library 9910, Get Model Score from Dropping Columns with Missing Values 18817, Elastic Net Regression 23386, One issue with yolo is that it is likely to contain more cells in its label grid that contain no objects than cells that do contain objects 12211, I hope it is now evident why I kept implementing a get features name method in the previous classes 1166, Most of these can be filled with None 6538, here i have merged some columns to just reduce complexity i have tried with all the columns but i didn t get this much accuracy which i am getting right now 5257, First of all we are going to train a baseline RF model 24059, We ll drop features with less than e importance you can change this threshold 18140, Here we average all the predictions and provide the final summary 21176, Displaying output of layer 4 31182, font color 5831bc face Comic Sans MS Before Scaling 24483, Training 10 folds for 10 epochs each strategy 3 improves validation score to 0 38925, Keras NN Model 8357, Survival rate by cabin 35939, Model evaluation 11816, SalePrice and Overall Quality 12616, Fare 12533, Cross validation of these models 19139, Model 2 input 784 ReLu 512 ReLu 128 sigmoid output 10 26288, Backward propagation with dropout 22774, Setting up the LSTM model 27175, Linear Regression 10140, let s try the same but using data with PCA applied 38893, Cost Function 34181, Interpreting CNN Model 2 6041, This code block finds best combinations for you 11399, time to deal with categorical features 13458, There are only 2 missing values for Embarked so we can just impute with the port where most people boarded 16739, embarked 36366, Train model on the full data and make final predictions 22655, Performance 2411, Numerical Imputing 36830, Accuracy 12532, Creating models 31529, Separating data based on data types 840, Model tuning and selection with GridSearchCV 13852, Numerical features 4299, Inference 15979, This feature is like the Name feature 11923, Feature Importance 28249, Little advanced visualization 4063, Clustering over the 3 most important principal components give us 80 of explained variance 33313, Plotting Decision Boundaries 38466, Columns with missing values either in train or in test 42837, We achieved good AUC on both training and cross validation 24471, Distribution regarding to target 18517, We perform a grayscale normalization to reduce the effect of illumination s differences 3982, logarithm the value of the house 32024, we can apply one hot encoding 33302, Model Selection 35396, Preprocessing by BoxCox 16379, Fare Band 28957, Some of those look quite good 26935, That s almost all we can train the classifier and evaluate it s performance 41343, Categorical Features 6226, saving files 2534, that we have checked the devices available we test them wth a simple computation 34054, Play with the parameters and do Cross Validation 37098, Outliers 25683, System with SOCIAL DISTANCING 6317, Multi Layer Perceptron 40622, Just a technical point add caching for the data pipeline 5326, Display spots of latitude and longitude 4184, Date and Time Engineering 35881, Sales data 36712, Different Tokenizers 26730, Plotting sales time series accross categories 27847, Top 20 2 gram words in sincere and insincere questions 41454, Feature Passenger Classes 35565, An important thing to note here is that this weird relationship between meta features and target does NOT extend to the test data generate predictions to demonstrate this 14204, Checking with some cross validation using stratified k fold 13301, We can now rank our evaluation of all the models to choose the best one for our problem 25848, Dealing with username 43270, Treinando o modelo com os dados de treino 17450, RandomForestClassifier 23558, These 37 records have living area greater than its total area 23196, Correlation among Base Models Predictions How base models predictions are correlated If base models predictions are weakly correlated with each other the ensemble likely to perform better On the other hand for a strong correlation of predictions among the base models the ensemble unlikely to perform better To sumarize diversity of predictions among the base models is inversely proportional to the ensemble accuracy make prediction for the test set 5485, For TotalBsmtSF only one outlier was there which was at index 1298 and same got deleted with GrLivArea 11680, get dummies creates a new columns for each of the options in Sex so that it creates a new columns for female called Sex female and new columns for male called Sex male which encodes if that row was male or female 37110, Code in python 1569, Cabin Number 2004, Again with the bottom two data points 16853, split 7299, Observation The distribution of SalePrice is unimodal in nature with a peak at 1500000 dollars 7853, Random Forests A Slight Improvement 12767, i normalize the test set as i did before and fill null values in the dataset as well 9027, Since there is only 1 null row with BsmtFinType2 and BsmtFinType2 is highly correlated with BsmtFinSF2 I got the BsmtFinSF2 value of the null row 24244, Name length 27909, Simple Imputer 34761, Fitting model on Training data 38715, Great so now the GPU is working and should speed up our computations 16874, Pclass Vs Survival 36597, step is loading the data and do pre processing and visualising it 1292, Ensembling 5 4 20571, Fare denotes the fare paid by a passenger 6070, MiscFeature ininfluencial drop it 19982, There are a lot of hyperparamater tuning when it comes to Keras such as 20473, Comparison of interval values with TARGET 1 and TARGET 0 13793, XGBoost 26345, We predict SalePrice column 37545, Data Preparation for Pytorch Model 26076, Chart of the number of digits in the data 41913, We can t really balance the size of our dataset by down sampling because almost all images are very large because of this we are going to resize our images instead 36775, Making Submission to Kaggle 35503, Evaluate the model 10452, Gradient Boosting Regression 35084, Comparing the images before and after applying the PCA 26399, There are twice as much men on board than women 7016, Evaluates the quality of the material on the exterior 3709, Lasso Regression 32797, Generate level 1 OOF predictions 15388, load training and test data 27097, The intermediate feature vector is the output of pool 3 or pool 4 and the global feature vector are fed as input to the attention layer 7218, Basement Null values 17855, we prepare the submission dataset and save it to the submission file 6229, Convert and create new features 14740, Dimensionality Reduction 14954, Extract title from name 40681, Show the distribution of distances of data samples from the most frequent template 2942, Separate Numerical and Categorical Features 34014, Monday 0 20078, Insights 25883, Histogram Plots of number of characters per each class 0 or 1 16622, Bayesian Optimization 21846, Training 23791, Modeling and Training 14334, One Hot Encoding 11443, 128187 Load and Read DataSets 8052, Advanced Regression Techniques 10660, Compare Model 37134, cross validation 23584, Submission 16666, This feature is from Konstantin s kernel 33876, Gradient Boosting Model 39975, Check seasonality of hour if we assume feature time is minute 19892, Getting the new train and test sets 10868, Transform the Name feature 3983, Check some Null s 14873, quite a few more people died than those who survived 4976, We can easily create a new feature called family size for Parch SibSp Himself Herself 15073, Embarked 12985, Train and test split