app.py

import streamlit as st
import streamlit.components.v1 as components
import pandas as pd
import pickle
import matplotlib.pyplot as plt
import pybanking
from pybanking.value_prediction import model_value_prediction
from pybanking.EDA import data_analysis 
import sklearn.metrics as metrics
from mlxtend.plotting import plot_confusion_matrix
import streamlit.components.v1 as components
from sklearn.model_selection import train_test_split
from PIL import Image
from pycaret.regression import pull,predict_model


st.set_page_config(page_title="Customer Value Prediction Model", layout="wide")

col1,col2 = st.columns([1,2])

with col1:
     image = Image.open('Shorthills.png')
     st.image(image)

with col2:
     st.title('Customer Value Prediction Model')

train_df= model_value_prediction.get_data()


option2 = st.selectbox(
     'Which dataset would you like to use for prediction?',
     ['Sample Dataset', 'Upload Custom']
     )

if option2 == 'Upload Custom':
     file = st.file_uploader("Choose a file")
     if file is not None:
          #read csv
          train_df = pd.read_csv(file)
     else:
          st.warning("you need to upload a csv file.")

st.subheader('This is the Selected Data')

st.dataframe(train_df.head(5))

analysis_class = data_analysis.Analysis()

option3 = st.selectbox(
     'Select Exploratory Data Analysis type',
     ['None', 'DataPrep', 'SweetViz', 'PandasProfiling']
     )

if option3 == 'SweetViz':
     res = analysis_class.sweetviz_analysis(train_df)
     res.show_html(filepath='SweetViz.html', open_browser=True, layout='widescreen', scale=None)
     HtmlFile = open('SweetViz.html', 'r', encoding='utf-8')
     source_code = HtmlFile.read()
     with st.expander("See Report"):
          components.html(source_code, height=600, scrolling=True)

elif option3 == 'DataPrep':
     res = analysis_class.dataprep_analysis(train_df)
     # res.show_browser()
     res.save('DataPrep.html')
     HtmlFile = open('DataPrep.html', 'r', encoding='utf-8')
     source_code = HtmlFile.read()
     with st.expander("See Report"):
          components.html(source_code, height=600, scrolling=True)
     
elif option3 == 'PandasProfiling':
     res = analysis_class.pandas_analysis(train_df)
     res.to_file("PandasProfiling.html")
     HtmlFile = open('PandasProfiling.html', 'r', encoding='utf-8')
     source_code = HtmlFile.read()
     with st.expander("See Report"):
          components.html(source_code, height=600, scrolling=True)

model_names = [
     "Logistic_Regression",
     "Support_Vector_Machine",
     "Support_Vector_Machine_Optimized",
     "Decision_Tree",
     "Neural_Network",
     "Random_Forest",
     "Pycaret_Best",
     "LGBM",
     "Lasso"
     ]


option = st.selectbox(
     'Select a model to be used',
     model_names
     )

tr_df = model_value_prediction.important_feat(train_df,option)
model = pickle.load(open(option+'.pkl', 'rb'))

st.write("Model Loaded :  ", option)

train_X,test_X,train_y,dev_X,val_X,dev_y,val_y,test_y= model_value_prediction.preprocess_inputs(tr_df,option)

model = model_value_prediction.train(tr_df,option)
y_pred = model_value_prediction.predict(test_X,model,option)

if option == "Pycaret_Best":
    predict_model(model)
    st.write("RMSLE Score:", pull()['RMSLE'][0])
else:
    st.write("RMSLE Score:", metrics.mean_squared_log_error(test_y, y_pred, squared=False))
    st.write("Poisson Score:", metrics.mean_tweedie_deviance(test_y, y_pred))