predict.py

import streamlit as st
from tensorflow.keras.models import load_model
import pandas as pd
import numpy as np
import pickle
from datetime import datetime , time

# Load All Files
with open('prep_pipeline.pkl','rb') as file1:
    prep = pickle.load(file1)

ann = load_model('churn_model.h5')

tanggal_waktu_sekarang = datetime.now().time()
waktu_sekarang = datetime.now().date()


def run():
  with st.form('key=Churn_prediction'):
      user_id = st.text_input('user_id:', '')
      age = st.number_input("age:" ,min_value=18 , max_value=100,step=1)
      gender = st.radio('gender', ('F','M'))
      region_category = st.radio('region_category', ('Town','City','Village'))
      membership_category = st.selectbox('membership_category', ('No Membership','Basic Membership','Silver Membership','Gold Membership','Platinum Membership','Premium Membership'), index=1)
      st.markdown('---')

      joining_date = st.date_input('joining_date', waktu_sekarang)
      joined_through_referral = st.radio('joined_through_referral',(True,False))
      preferred_offer_types = st.selectbox('preferred_offer_types',('Gift Vouchers/Coupons','Credit/Debit Card Offers','Without Offers'))
      medium_of_operation = st.selectbox('medium_of_operation',('Desktop','Smartphone','Both'))
      internet_option = st.radio('internet_option',('Wi-Fi','Mobile_Data','Fiber_Optic'))
      last_visit_time = st.time_input('last_visit_time', tanggal_waktu_sekarang)
      days_since_last_login = st.number_input('days_since_last_login', min_value=0 , max_value=1000,step=1)
      avg_time_spent = st.number_input('avg_time_spent' , min_value=0 , max_value=3300,step=1)
      avg_transaction_value = st.number_input('avg_transaction_value', min_value=500 , max_value=100000,step=5)
      avg_frequency_login_days = st.number_input('avg_frequency_login_days', min_value=0 , max_value=75,step=1)
      points_in_wallet = st.number_input('points_in_wallet',min_value=0 , max_value=2100,step=5)
      used_special_discount = st.radio('used_special_discount', ('Yes','No'))
      offer_application_preference = st.radio('offer_application_preference',('Yes','No'))
      past_complaint = st.radio('past_complaint',('Yes','No'))
      complaint_status = st.selectbox('complaint_status', ('Not Applicable' , 'Unsolved','Solved','Solved in Follow-up','No Information Available'))
      feedback = st.selectbox('feedback', ('Poor Product Quality','No reason specified','Too many ads','Poor Website','Poor Customer Service','Reasonable Price','User Friendly Website','Products always in Stock','Quality Customer Care'))
      submitted = st.form_submit_button('Predict')

  data_inf = {
      'user_id': user_id,
      'age': age,
      'gender': gender,
      'region_category': region_category,
      'membership_category': membership_category,
      'joining_date': joining_date,
      'joined_through_referral': joined_through_referral,
      'preferred_offer_types': preferred_offer_types,
      'medium_of_operation': medium_of_operation,
      'internet_option': internet_option,
      'last_visit_time': last_visit_time,
      'days_since_last_login': days_since_last_login,
      'avg_time_spent' : avg_time_spent,
      'avg_transaction_value' : avg_transaction_value,
      'avg_frequency_login_days' : avg_frequency_login_days,
      'points_in_wallet' : points_in_wallet,
      'used_special_discount' : used_special_discount,
      'offer_application_preference' : offer_application_preference,
      'past_complaint' : past_complaint,
      'complaint_status' : complaint_status,
      'feedback' : feedback
  }
  
  data_inf = pd.DataFrame([data_inf])
  positive_words = ['reasonable', 'friendly', 'always', 'stock','care','price','user','products','in']
  negative_words = ['poor', 'too', 'many', 'terrible','product quality','ads','service']
  def classify_sentiment(text):
        words = text.lower().split()
        num_positive = sum(1 for word in words if word in positive_words)
        num_negative = sum(1 for word in words if word in negative_words)
    
        if num_positive > num_negative:
            return 'positif'
        elif num_positive < num_negative:
            return 'negatif'
        else:
            return 'netral'
# Menerapkan fungsi klasifikasi sentimen menggunakan apply pada DataFrame
  data_inf['sentimen'] = data_inf['feedback'].apply(classify_sentiment)

  final_data = prep.transform(data_inf)


  if submitted:
    y_pred_inf = ann.predict(final_data)
    y_pred_inf = np.where(y_pred_inf >= 0.5 , 'Churn' , 'Not Churn')
    st.write('Price of the car : ', str(y_pred_inf))


if __name__ == '__main__':
    run()