app.py

import streamlit as st
import pandas as pd
import joblib
import numpy as np

# 1. Load the saved artifacts
model = joblib.load('churn_model.pkl')
scaler = joblib.load('scaler.pkl')
model_columns = joblib.load('features.pkl')

# 2. App Title and Description
st.title("📊 Customer Retention Engine")
st.write("Enter customer details to predict the probability of churn.")

# 3. Create Input Fields for the User
# We focus on the most important features we found in Phase 3
st.sidebar.header("Customer Input")

tenure = st.sidebar.slider("Tenure (Months)", 0, 72, 12)
monthly_charges = st.sidebar.number_input("Monthly Charges ($)", min_value=18.0, max_value=120.0, value=70.0)
total_charges = st.sidebar.number_input("Total Charges ($)", min_value=18.0, max_value=8000.0, value=tenure * 70.0)

# Categorical Inputs (Dropdowns)
contract = st.sidebar.selectbox("Contract Type", ["Month-to-month", "One year", "Two year"])
internet_service = st.sidebar.selectbox("Internet Service", ["DSL", "Fiber optic", "No"])
payment_method = st.sidebar.selectbox("Payment Method", ["Electronic check", "Mailed check", "Bank transfer", "Credit card"])
online_security = st.sidebar.selectbox("Online Security", ["Yes", "No", "No internet service"])

# 4. Logic to convert inputs into the format the model expects
def preprocess_input():
    # Create a dict with the numerical inputs
    input_data = {
        'tenure': tenure,
        'MonthlyCharges': monthly_charges,
        'TotalCharges': total_charges
    }
    
    # Initialize a dataframe with 0s for all model columns
    df_input = pd.DataFrame(columns=model_columns)
    df_input.loc[0] = 0  # Initialize row with 0
    
    # Map Numerical Inputs
    df_input['tenure'] = input_data['tenure']
    df_input['MonthlyCharges'] = input_data['MonthlyCharges']
    df_input['TotalCharges'] = input_data['TotalCharges']
    
    # Handle Categorical (Manual One-Hot Encoding Simulation)
    # We check if the column exists in our trained features, then set it to 1
    
    # Contract
    if contract == "One year" and 'Contract_One year' in df_input.columns:
        df_input['Contract_One year'] = 1
    elif contract == "Two year" and 'Contract_Two year' in df_input.columns:
        df_input['Contract_Two year'] = 1
    # Note: Month-to-month is the baseline (all 0s), so we don't need to set anything!
    
    # Internet Service
    if internet_service == "Fiber optic" and 'InternetService_Fiber optic' in df_input.columns:
        df_input['InternetService_Fiber optic'] = 1
    elif internet_service == "No" and 'InternetService_No' in df_input.columns:
        df_input['InternetService_No'] = 1
        
    # Payment Method (Example of mapping specific selection)
    if payment_method == "Electronic check" and 'PaymentMethod_Electronic check' in df_input.columns:
        df_input['PaymentMethod_Electronic check'] = 1
        
    # Online Security
    if online_security == "Yes" and 'OnlineSecurity_Yes' in df_input.columns:
        df_input['OnlineSecurity_Yes'] = 1
        
    return df_input

# 5. Prediction Button
if st.button("Predict Churn Risk"):
    user_input = preprocess_input()
    
    # Scale the numerical columns (using the loaded scaler)
    cols_to_scale = ['tenure', 'MonthlyCharges', 'TotalCharges']
    user_input[cols_to_scale] = scaler.transform(user_input[cols_to_scale])
    
    # Predict
    prediction = model.predict(user_input)
    probability = model.predict_proba(user_input)[:, 1][0]
    
    st.subheader("Results:")
    
    if prediction[0] == 1:
        st.error(f"⚠️ High Risk of Churn! Probability: {probability:.2%}")
        st.write("Suggestion: Offer this customer a 1-year contract discount immediately.")
    else:
        st.success(f"✅ Low Risk of Churn. Probability: {probability:.2%}")
        st.write("Suggestion: Customer is stable. No immediate action needed.")