Spaces:
Sleeping
Sleeping
| import streamlit as st | |
| import joblib | |
| import numpy as np | |
| # input = ['clarity_I1', 'clarity_IF', 'clarity_SI1', 'clarity_SI2', 'clarity_VS1', | |
| # 'clarity_VS2', 'clarity_VVS1', 'clarity_VVS2', 'cut_Fair', 'cut_Good', | |
| # 'cut_Ideal', 'cut_Premium', 'cut_Very Good', 'color_D', 'color_E', | |
| # 'color_F', 'color_G', 'color_H', 'color_I', 'color_J', 'carat', 'depth', | |
| # 'table', 'x', 'y', 'z'] | |
| # carat : 0.20 -> 3.24 | |
| # depth : 43 - 80 | |
| # table : 50 - 80 | |
| # x : 3.73 - 9.54 | |
| # y : 3.68 - 9.46 | |
| # z : 1.07 - 5.98 | |
| # Output: Price | |
| model = joblib.load('./DiamondsPrices_model.pkl') | |
| def predict_price(clarity, cut, color, carat, depth, table, x, y, z): | |
| columns = ['clarity_I1', 'clarity_IF', 'clarity_SI1', 'clarity_SI2', 'clarity_VS1', | |
| 'clarity_VS2', 'clarity_VVS1', 'clarity_VVS2', 'cut_Fair', 'cut_Good', | |
| 'cut_Ideal', 'cut_Premium', 'cut_Very Good', 'color_D', 'color_E', | |
| 'color_F', 'color_G', 'color_H', 'color_I', 'color_J', 'carat', 'depth', | |
| 'table', 'x', 'y', 'z'] | |
| x_input = np.zeros(len(columns)) | |
| x_input = x_input.tolist() | |
| x_input[20] = carat | |
| x_input[21] = depth | |
| x_input[22] = table | |
| x_input[23] = x | |
| x_input[24] = y | |
| x_input[25] = z | |
| clarity_index = columns.index(clarity) | |
| x_input[clarity_index] = 1 | |
| cut_index = columns.index(cut) | |
| x_input[cut_index] = 1 | |
| color_index = columns.index(color) | |
| x_input[color_index] = 1 | |
| x_input = np.array(x_input).reshape(1, -1) | |
| price = model.predict(x_input) | |
| return price[0] | |
| st.image("logo.png", width=64) | |
| st.title('Diamonds Prices Prediction') | |
| st.subheader( | |
| 'Please read about the dataset that we used to build the model on the side menu. Model on: XGBRegressor: 0.9812 accuracy.")') | |
| col1, col2, col3 = st.columns(3) | |
| with col1: | |
| st.write('Diamond Clarity') | |
| clarity = st.selectbox( | |
| 'Clarity', ['clarity_I1', 'clarity_IF', 'clarity_SI1', 'clarity_SI2', 'clarity_VS1', | |
| 'clarity_VS2', 'clarity_VVS1', 'clarity_VVS2']) | |
| cut = st.selectbox( | |
| 'Cut', ['cut_Fair', 'cut_Good', | |
| 'cut_Ideal', 'cut_Premium', 'cut_Very Good']) | |
| color = st.selectbox( | |
| 'Color', ['color_D', 'color_E', | |
| 'color_F', 'color_G', 'color_H', 'color_I', 'color_J']) | |
| with col2: | |
| st.write('Carat, Depth, & Table') | |
| carat = st.number_input('Carat', min_value=0.20, | |
| max_value=3.50, value=1.20) | |
| depth = st.number_input('Depth', min_value=52.0, | |
| max_value=70.0, value=55.0) | |
| table = st.number_input('Table', min_value=52.0, | |
| max_value=70.0, value=52.0) | |
| with col3: | |
| st.write('Diamond Dimensions') | |
| x = st.number_input('X', min_value=3.0, max_value=10.0, value=3.20) | |
| y = st.number_input('Y', min_value=3.0, max_value=10.0, value=3.73) | |
| z = st.number_input('Z', min_value=1.0, max_value=6.0, value=1.5) | |
| if st.button('Predict'): | |
| price = predict_price(clarity, cut, color, carat, depth, table, x, y, z) | |
| st.subheader('The price of the diamond is Around:') | |
| st.subheader(f'${price:.2f}') | |
| with st.sidebar: | |
| st.image("logo.png", width=64) | |
| st.title('About Diamonds Prices Prediction App.') | |
| st.write("This App is built using the dataset from kaggle by: VITTORIO GIATTI.") | |
| dataset_url = 'https://www.kaggle.com/datasets/vittoriogiatti/diamondprices' | |
| st.markdown('Dataset [link](%s)' % dataset_url) | |
| st.subheader("About the DataSet") | |
| st.write('Each record is a random diamond 💎 💍 with its own characteristics. It should be easy to create a model to predict its price on the market basing on objective variables or try to cluster the observations avoiding the use of clarity, which is the "pureness" of the stone.') | |
| st.write("We tried from scikit-learn The following algorithms for regression:") | |
| st.write("1. Linear Regression: 0.8964 accuracy.") | |
| st.write("2. Lasso: 0.8961 accuracy.") | |
| st.write("3. DecisionTreeRegressor: 0.8649 accuracy.") | |
| st.write("We Also we tried from XGBoost:") | |
| st.write("1. XGBRegressor: 0.9301 accuracy.") | |
| st.write("So we saved the model of XGBRegressor algorithm in a pickle file and we are using it here.") | |
| st.subheader( | |
| "Updated the App with the following changes:") | |
| st.write("1. Did outliers removing with accurate values.") | |
| st.write("2. So I retained around ~ 53700 rows, while the original dataset has 53940 rows. But the first version took around ~1000 as outliers.") | |
| st.write("3. I also retained categorigcal columns (cut, color and clarity) and used them in the model.") | |
| st.write("The result was :") | |
| st.write("scikit-learn algorithms:") | |
| st.write("1. Linear Regression: 0.9227 accuracy.") | |
| st.write("2. Lasso: 0.9224 accuracy.") | |
| st.write("3. DecisionTreeRegressor: 0.9674 accuracy.") | |
| st.write("XGBoost:") | |
| st.write("1. XGBRegressor: 0.9812 accuracy.") | |
| st.write("So we saved the model of XGBRegressor algorithm in a pickle file and we are using it here.") | |
| st.write( | |
| "New code is in files : video12-B.py and video13-B.py in github, link below:") | |
| github_url = 'https://github.com/shangab/PracticalAI/tree/main/STAT_ML_DS' | |
| st.markdown('Github [link](%s)' % github_url) | |