app.py

import pandas as pd
import numpy as np
from sklearn.preprocessing import StandardScaler, OneHotEncoder, LabelEncoder, OrdinalEncoder
from sklearn.linear_model import LinearRegression, LogisticRegression
from sklearn.ensemble import GradientBoostingClassifier
import xgboost
from sklearn.compose import ColumnTransformer
#import pickle
from sklearn.pipeline import Pipeline
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report, r2_score
import streamlit as st
import shap
import matplotlib as mt

def train(data=None,problem="Regression",model="LinearRegression",label=None):

    df = pd.read_csv(data)

    target = df[label].copy()
    features = df.drop(label, axis=1)

    X_train,X_test,y_train,y_test = train_test_split(features,target,test_size=0.20,random_state=42,shuffle=True,stratify=target)

    num_features = []
    cat_features = []
    cols = list(features.columns)
    for i in cols:
        if df[i].dtypes == "object":
            cat_features.append(i)
        else:
            num_features.append(i)

    if problem == "Regression":
        if cat_features[0]!="":

            trf = ColumnTransformer([("num_trf",StandardScaler(),num_features),
                                 ("cat_trf",OneHotEncoder(sparse_output=False),cat_features)])
        else:
            trf = ColumnTransformer([("num_trf",StandardScaler(),num_features)])
            
        
        
        final_pipe = Pipeline([("transformers",trf),("reg_model",LinearRegression())])

        final_pipe.fit(X_train,y_train)

        #model = pickle.dump(final_pipe,open("regression_model","wb"))

        #y_hat = model.predict(X_train)

        return final_pipe, X_train,X_test,y_train,y_test
    if problem == "Classification":
        if model == "GradientBoosting":

            trf = ColumnTransformer([("num_trf",StandardScaler(),num_features),
                                    ("cat_trf",OneHotEncoder(),cat_features)])
            
            
            lbl_encd = LabelEncoder()

            lbl_encd.fit(y_train)
            y_train_trf = lbl_encd.transform(y_train)

            y_test_trf = lbl_encd.fit(y_test)
            
            final_pipe = Pipeline([("transformers",trf),("clf_model",GradientBoostingClassifier(random_state=42))])

            final_pipe.fit(X_train,y_train_trf)
            #file = open("model")
            #model = pickle.dump(final_pipe,("","wb"))

            return final_pipe, X_train,X_test,y_train_trf,y_test_trf
        elif model == "LogisticRegression":
            trf = ColumnTransformer([("num_trf",StandardScaler(),num_features),
                                    ("cat_trf",OneHotEncoder(),cat_features)])
            
            
            lbl_encd = LabelEncoder()

            lbl_encd.fit(y_train)
            y_train_trf = lbl_encd.transform(y_train)

            y_test_trf = lbl_encd.fit(y_test)
            
            final_pipe = Pipeline([("transformers",trf),("clf_model",LogisticRegression(random_state=42))])

            final_pipe.fit(X_train,y_train_trf)
            #file = open("model")
            #model = pickle.dump(final_pipe,("","wb"))

            return final_pipe, X_train,X_test,y_train_trf,y_test_trf


def predict(model=None,x=None):

    #m = pickle.load(open(model,"rb"))
    y_hat = model.predict(x)

    return y_hat

def evaluate(y_true,y_pred, problem="Regression"):

    if problem == "Regression":
        metric = r2_score(y_true,y_pred)
        return metric
    else:
        metric = classification_report(y_true,y_pred,output_dict=True)
        met_df = pd.DataFrame(metric).transpose()
        file = met_df.to_csv().encode('utf-8')

        return file

st.title("No Code Machine Learning Studio: ")

st.image(image="https://www.silvertouchtech.co.uk/wp-content/uploads/2020/05/ai-banner.jpg")
st.write("Drag & Drop Portal for Machine Learing")
prob_type = st.selectbox(label="Please select your ML problem type: ",options=("Regression","Classification"))

train_data = st.file_uploader(label="Please upload your training dataset",type=["csv"])

if prob_type == "Classification":

    model = st.selectbox(label="Plase Select your classification model: ", options=("GradientBoosting","LogisticRegression"))
else:
    model = "LinearRegression"


def explain(model="LinearRegression",train_data=None,test_data=None):
    explainer = shap.LinearExplainer(model,train_data,feature_dependence=False)
    shap_values = explainer.shap_values(test_data)

    shap.summary_plot(shap_values,test_data,plot_type="violin",show=False)
    mt.pyplot.gcf().axes[-1].set_box_aspect(10)


y = st.text_input("Please write your target column name:  ")
#num_f = st.text_input("Please write your numerical feature names(separted by ","): ").split(",")
#cat_f = st.text_input("Please write your categorical feature names(separted by ","): ").split(",")

if st.button("Train"):
    #if cat_f[0]!="":
    model_, X_train,X_test,y_train,y_test = train(data=train_data,problem=prob_type,model=model, label=y)
    #else:
    #model_, X_train,X_test,y_train,y_test = train(data=train_data,problem=prob_type,model=model, label=y,num_features=num_f,cat_features=cat_f)

    y_hat_train = predict(model_,X_train)
    y_hat_test = predict(model_,X_test)

    if prob_type == "Classification":
        st.write("Classification report of training set: ")
        report = evaluate(y_train,y_hat_train,prob_type)

        st.download_button(label="Click here to download the report",data=report, mime="text/csv")
        st.write("Classification report of testing dataset: ")
        report_test = evaluate(y_train,y_hat_train,prob_type)
        st.download_button(key="test",label="Click here to download the report",data=report_test, mime="text/csv")

    else:
        st.write("r2 score on training set: ")
        st.write(evaluate(y_train,y_hat_train))
        st.write("r2 score on test set: ")

        st.write(evaluate(y_test,y_hat_test,prob_type))
        #explain(model_.named_steps["reg_model"],X_train,X_test)