app.py

import streamlit as st
import pandas as pd
from joblib import load

# Load the trained model and scaler
model = load('loandefaulter.joblib')
scaler = load('scaler.joblib')

# Define numerical features for scaling (only those that were used during training)
num_features = [
    'loan_amnt', 'int_rate', 'installment', 'annual_inc', 'dti',
    'revol_bal', 'revol_util', 'total_acc', 'mort_acc'
]

# Create the Streamlit app
st.set_page_config(page_title='Loan Default Prediction', layout='wide')

# App title and description
st.markdown("""
    <style>
    .title { font-size: 36px; font-weight: bold; color: #2E86C1; }
    .description { font-size: 20px; color: #34495E; }
    .input-container { margin-top: 20px; }
    .slider-container { margin: 10px 0; }
    </style>
    <div class="title">Loan Default Prediction</div>
    <div class="description">Enter the loan details below to get a prediction on whether the loan will be defaulted.</div>
""", unsafe_allow_html=True)

# Input fields with sliders
loan_amnt = st.slider('Loan Amount', min_value=0.0, max_value=1000000.0, step=1000.0, value=10000.0)
int_rate = st.slider('Interest Rate (%)', min_value=0.0, max_value=30.0, step=0.1, value=5.0)
installment = st.slider('EMI Amount', min_value=0.0, max_value=10000.0, step=10.0, value=200.0)
annual_inc = st.slider('Annual Income', min_value=0.0, max_value=1000000.0, step=1000.0, value=50000.0)

# CIBIL score input as a text field
cibil_score = st.text_input('CIBIL Score (Enter a number between 300 and 900)', value='700')

# Convert the CIBIL score input to a numeric value
try:
    cibil_score = int(cibil_score)
except ValueError:
    st.error("Please enter a valid number for CIBIL Score.")

# Set default values for the missing features
dti = 0.0  # Example default value for DTI
revol_bal = 0.0  # Example default value for Revolving Balance
revol_util = 0.0  # Example default value for Revolving Utilization
total_acc = 0  # Example default value for Total Accounts
mort_acc = 0  # Example default value for Mortgage Accounts

loan_amnt_by_income = loan_amnt / (annual_inc + 1)

# Create a DataFrame for the input
input_data = pd.DataFrame({
    'loan_amnt': [loan_amnt],
    'int_rate': [int_rate],
    'installment': [installment],
    'annual_inc': [annual_inc],
    'dti': [dti],
    'revol_bal': [revol_bal],
    'revol_util': [revol_util],
    'total_acc': [total_acc],
    'mort_acc': [mort_acc]
})

# Scale the numerical features that were used to fit the scaler
input_data[num_features] = scaler.transform(input_data[num_features])

# Add the additional feature (not part of scaling)
input_data['loan_amnt_by_income'] = [loan_amnt_by_income]


input_data['cibil_score'] = cibil_score  
input_data = input_data[num_features + ['loan_amnt_by_income']]  

# Predict using the model
if st.button('Predict'):
    if 300 <= cibil_score <= 900:  # Ensure CIBIL score is within the valid range
        prediction = model.predict(input_data)
        result = "Defaulted" if prediction[0] == 1 else "Not Defaulted"
        color = "red" if prediction[0] == 1 else "green"
        st.markdown(f"""
            <div style="font-size: 24px; color: {color}; font-weight: bold;">Prediction: {result}</div>
        """, unsafe_allow_html=True)
    else:
        st.error("Please enter a CIBIL Score between 300 and 900.")