CreditWise — AI-Powered Loan Intelligence System

CreditWise is a full-stack Django web application that integrates Machine Learning to automate loan approval decisions. The system analyzes applicant financial details and predicts whether a loan application should be Approved or Rejected, along with the probability of approval.

The prediction engine uses a Logistic Regression model trained on 975,000+ realistic Indian financial records, enabling data-driven decision making similar to real banking risk assessment systems.

Features

• Loan Approval Predictor — Enter your financial profile and get real-time approval probability with animated visual feedback
• Key Factor Analysis — See how your credit score, DTI ratio, collateral, and existing loans affect your chances
• Improvement Tips — Personalized suggestions to improve approval odds
• EMI Calculator — Calculate monthly EMI with three interest methods: Reducing Balance, Flat Rate, and Compound Interest
• Amortization Schedule — Year-by-year repayment breakdown with donut chart visualization
• DTI Auto-Calculator — Automatically calculates your Debt-to-Income ratio from monthly income and debt inputs

---

Tech Stack

Layer	Technology
Backend	Python, Django
ML Model	Scikit-learn (Logistic Regression)
Data Processing	Pandas, NumPy
Frontend	HTML, CSS, Vanilla JavaScript
Styling	Custom dark luxury theme
Model Persistence	Joblib

---

Machine Learning Model

• Algorithm — Logistic Regression
• Training Data — 975,800 rows of synthetic Indian loan data
• Accuracy — 83.1%
• Precision — 74.8%
• Recall — 64.6%
• F1 Score — 69.3%

Features Used (27 total)

• Applicant Income, Co-applicant Income, Age, Dependents
• Credit Score (squared), DTI Ratio (squared), Collateral Ratio
• Savings, Loan Amount, Loan Term, Existing Loans
• Employment Status, Employer Category, Education Level
• Marital Status, Gender, Loan Purpose, Property Area

Key Correlations with Approval

Feature	Correlation
Credit Score²	+0.38
Employment (Salaried)	+0.27
Collateral Ratio	+0.16
DTI Ratio²	-0.31
Existing Loans	-0.19
Loan Amount	-0.13

---

Project Structure

CreditWiseLoanSystem/
├── ML/
│   ├── Data/
│   │   ├── loan_approval_data.csv
│   │   └── Processed_loan_approval_data.csv
│   └── Train and Test Model/
│       └── final_model.py
├── model/
│   ├── loan_model.pkl
│   └── scaler.pkl
└── web/
    ├── manage.py
    ├── settings.py
    ├── urls.py
    ├── views.py
    ├── wsgi.py
    ├── templates/
    │   └── index.html
    └── static/
        ├── style.css
        └── script.js

---

Installation & Setup

Prerequisites

• Python 3.8+
• Anaconda (recommended)

Steps

1. Clone the repository

git clone https://github.com/yourusername/CreditWiseLoanSystem.git
cd CreditWiseLoanSystem

2. Install dependencies

pip install django scikit-learn pandas numpy joblib whitenoise gunicorn

3. Run the development server

cd web
python manage.py runserver

4. Open in browser

http://127.0.0.1:8000

---

Retraining the Model

If you want to retrain with new data:

# Step 1 — Run the preprocessing notebook
# Open ML/data_Preprocessing.ipynb and run all cells
# This generates ML/Data/Processed_loan_approval_data.csv

# Step 2 — Train the model
python "ML/Train and Test Model/final_model.py"

# Step 3 — Restart the server
cd web
python manage.py runserver

---

Input Ranges

The model accepts real Indian rupee values:

Field	Range
Monthly Income	₹10,000 – ₹1,00,00,000
Loan Amount	₹10,000 – ₹50,00,00,000
Savings	₹1,000 – ₹10,00,00,000
Collateral Value	₹0 – ₹1,00,00,00,000
Credit Score	300 – 900
DTI Ratio	0.05 – 0.90

---

📊 Dataset

The dataset used to train this model is publicly available on Kaggle:

🔗 CreditWise Loan Approval Dataset — Kaggle

• 1,000,000 synthetic Indian loan applications
• Realistic Indian rupee ranges (₹10,000 – ₹1,00,00,000 income)
• 13 features including credit score, DTI ratio, collateral, employment status
• Target: Loan_Approved (1 = Approved, 0 = Rejected)

---

Deployment

This project is configured for deployment on Render.

Environment Variables

Key	Value
SECRET_KEY	Your secret key
DEBUG	False
ALLOWED_HOSTS	your-app-name.onrender.com

Build Command

pip install -r requirements.txt && cd web && python manage.py collectstatic --noinput

Start Command

cd web && gunicorn wsgi:application --bind 0.0.0.0:$PORT

---

Author

Mayank — Built as a full-stack ML project combining data science, Django backend, and frontend UI design.

---

License

This project is for educational purposes.