app.py

import streamlit as st
import numpy as np
import tensorflow as tf
from sklearn.preprocessing import StandardScaler, LabelEncoder, OneHotEncoder
import pandas as pd
import pickle
import matplotlib.pyplot as plt

# Load the trained model
model = tf.keras.models.load_model('model.h5')

# Load the encoders and scaler
with open('label_encoder_gender.pkl', 'rb') as file:
    label_encoder_gender = pickle.load(file)

with open('onehot_encoder_geo.pkl', 'rb') as file:
    onehot_encoder_geo = pickle.load(file)

with open('scaler.pkl', 'rb') as file:
    scaler = pickle.load(file)

# Set Streamlit page configuration
st.set_page_config(page_title="Churn Predictor", page_icon='Fardeen NB - Logo.png', layout="centered", initial_sidebar_state="auto", menu_items=None)

# App title
st.title('Customer Churn Prediction Project')

# Create two columns for input layout
col1, spacer, col2 = st.columns([1, 0.2, 1])

# Input elements arranged side by side in the two columns
with col1:
    geography = st.selectbox('Geography', onehot_encoder_geo.categories_[0])
    gender = st.selectbox('Gender', label_encoder_gender.classes_)
    age = st.slider('Age', 18, 92)
    balance = st.number_input('Balance')
    credit_score = st.number_input('Credit Score')

with col2:
    estimated_salary = st.number_input('Estimated Salary')
    tenure = st.slider('Tenure', 0, 10)
    num_of_products = st.slider('Number of Products', 1, 4)
    has_cr_card = st.selectbox('Has Credit Card', [0, 1])
    is_active_member = st.selectbox('Is Active Member', [0, 1])

# Prepare the input data for prediction
input_data = pd.DataFrame({
    'CreditScore': [credit_score],
    'Gender': [label_encoder_gender.transform([gender])[0]],
    'Age': [age],
    'Tenure': [tenure],
    'Balance': [balance],
    'NumOfProducts': [num_of_products],
    'HasCrCard': [has_cr_card],
    'IsActiveMember': [is_active_member],
    'EstimatedSalary': [estimated_salary]
})

# One-hot encode 'Geography'
geo_encoded = onehot_encoder_geo.transform([[geography]]).toarray()
geo_encoded_df = pd.DataFrame(geo_encoded, columns=onehot_encoder_geo.get_feature_names_out(['Geography']))

# Combine one-hot encoded columns with input data
input_data = pd.concat([input_data.reset_index(drop=True), geo_encoded_df], axis=1)

# Scale the input data
input_data_scaled = scaler.transform(input_data)

# Predict churn probability
prediction = model.predict(input_data_scaled)
prediction_proba = prediction[0][0]

# Display churn probability
st.write( f'Churn Probability: {prediction_proba * 100:.2f}%')

# Churn likelihood output
if prediction_proba > 0.3:
    st.markdown("## The customer is likely to churn.")
else:
    st.markdown("## The customer is not likely to churn.")

# Plotting churn probability
# Create a plot using matplotlib
fig, ax = plt.subplots()

# Set y-axis limit to 100 for distinction
ax.set_ylim(0, 100)

# Set background color to black
fig.patch.set_facecolor('black')
ax.set_facecolor('black')

# Plot the churn percentage as a bar chart with rounded corners
churn_percentage = prediction_proba * 100
bars = ax.bar(['Churn Probability'], [churn_percentage], color='darkblue', edgecolor='darkblue', linewidth=0, width=0.5)

# Adding labels and title with white font
ax.set_ylabel('Percentage (%)', color='white')
ax.set_title('Customer Churn Probability', color='white')

# Customize tick colors
ax.tick_params(axis='x', colors='white')
ax.tick_params(axis='y', colors='white')

# Show percentage on the bar with white font and padding on top
for i, v in enumerate([churn_percentage]):
    ax.text(i, v + 3, f"{v:.2f}%", ha='center', color='white')

# Add watermark "© Copyrighted to https://fardeen.net ™" in pure white
watermark_text = "© 2024 Fardeen NB. All Rights Reserved. https://fardeen.net ™"
fig.text(0.5, 0.001, watermark_text, fontsize=10, color='white', ha='center', alpha=0.8)

# Display the plot in Streamlit
st.pyplot(fig)