https://github.com/rahulrmcoder/smart-lending-ally
Smart Lending Ally is an AI-driven loan recommendation platform that helps borrowers find optimal lending options across 10 major Indian banks using dual-model classification (XGBoost + Random Forest) with 98.33% accuracy.
https://github.com/rahulrmcoder/smart-lending-ally
Last synced: 9 months ago
JSON representation
Smart Lending Ally is an AI-driven loan recommendation platform that helps borrowers find optimal lending options across 10 major Indian banks using dual-model classification (XGBoost + Random Forest) with 98.33% accuracy.
- Host: GitHub
- URL: https://github.com/rahulrmcoder/smart-lending-ally
- Owner: RahulRmCoder
- License: mit
- Created: 2025-06-25T14:55:55.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2025-06-25T14:59:02.000Z (about 1 year ago)
- Last Synced: 2025-06-25T15:49:44.307Z (about 1 year ago)
- Language: Jupyter Notebook
- Size: 0 Bytes
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Smart Lending Ally: AI-Powered Loan Recommendation System
## 🎯 Overview
Smart Lending Ally is an innovative AI-driven loan recommendation platform that revolutionizes how borrowers find the most suitable lending options in India's complex financial landscape. The system utilizes a unique dual-model classification approach combining XGBoost and Random Forest algorithms to predict loan eligibility across 10 prominent Indian banks with an impressive 98.33% average accuracy.
### Key Features
- **🤖 Dual-Model Classification**: Intelligently selects between XGBoost and Random Forest models based on performance metrics
- **🎯 Multi-Criteria Suitability Scoring**: Goes beyond simple eligibility to evaluate interest rates, processing times, and affordability
- **📊 Dynamic Weighting System**: Customized recommendations for 5 loan types (Home, Car, Education, Personal, Gold)
- **🏦 Multi-Bank Coverage**: Supports 10 major Indian banks including SBI, HDFC, ICICI, Axis Bank, and more
- **📈 Comprehensive Visualization**: User-friendly charts and explanations for transparent decision-making
- **🔍 Explainable AI**: Natural language explanations for all recommendations
## 🏗️ System Architecture
The Smart Lending Ally system follows a sophisticated pipeline:
1. **Data Preprocessing**: Feature engineering with financial ratios (DTI, LTI) and categorical encoding
2. **Dual-Model Training**: Comparative selection between Random Forest and XGBoost for each bank
3. **Suitability Scoring**: Multi-criteria evaluation using weighted factors
4. **Recommendation Generation**: Ranked suggestions with detailed explanations
5. **Visualization**: Interactive charts and comprehensive reports
## 📊 Dataset
The system uses a comprehensive dataset featuring:
- **2,000 synthetic loan application profiles** representing realistic Indian financial scenarios
- **25 features per profile** including demographics, financial details, and loan specifics
- **10 major Indian banks** with specific eligibility criteria
- **5 loan categories**: Home, Car, Education, Personal, and Gold loans
- **Income range**: ₹15,000 to ₹200,000 monthly
- **CIBIL scores**: 300 to 900 range
### Supported Banks
| Bank | Accuracy | Model Used |
|------|----------|------------|
| IndusInd Bank | 99.25% | XGBoost |
| ICICI Bank | 98.75% | XGBoost |
| Punjab National Bank | 98.50% | XGBoost |
| Kotak Mahindra Bank | 98.50% | XGBoost |
| HDFC Bank | 98.25% | XGBoost |
| Axis Bank | 98.25% | XGBoost |
| State Bank of India | 97.75% | XGBoost |
| Union Bank of India | 97.50% | XGBoost |
| Bank of Baroda | 97.50% | XGBoost |
| Canara Bank | 96.75% | XGBoost |
## 🚀 Getting Started
### Prerequisites
- Python 3.11 or higher
- 16 GB RAM (recommended)
- Intel Core i7 processor or equivalent
### Installation
1. Clone the repository:
```bash
git clone https://github.com/yourusername/smart-lending-ally.git
cd smart-lending-ally
```
2. Install required dependencies:
```bash
pip install scikit-learn xgboost pandas numpy matplotlib seaborn plotly jupyter
```
3. Launch Jupyter Notebook:
```bash
jupyter notebook
```
4. Open the main notebook:
```bash
smart_lending_ally.ipynb
```
### Quick Start
1. **Open the Jupyter Notebook**: Launch `smart_lending_ally.ipynb` in your Jupyter environment
2. **Run all cells**: Execute the notebook cells sequentially to:
- Load and preprocess the synthetic dataset
- Train the dual-model classification system
- Generate loan recommendations
3. **Customize input profiles**: Modify the applicant profile in the notebook:
```python
# Example profile (modify in the notebook)
profile = {
'age': 42,
'monthly_income': 150000,
'cibil_score': 810,
'loan_type': 'Home',
'loan_amount': 8000000,
'employment_years': 15
# ... other features
}
```
4. **View results**: The notebook will generate:
- Eligibility predictions for all 10 banks
- Suitability scores and rankings
- Interactive visualizations
- Detailed recommendations with explanations
## 📈 Performance Metrics
### Model Performance
- **Average Accuracy**: 98.33% across all banks
- **Recommendation Precision**: 96.7% when validated against expert evaluations
- **Processing Time**: Real-time recommendations
- **Scalability**: Handles diverse user profiles effectively
### Key Improvements Over Existing Systems
- **20-28% higher accuracy** compared to single-algorithm systems (70-87%)
- **Comprehensive suitability scoring** beyond basic eligibility
- **Transparent explanations** for all recommendations
- **Dynamic weighting** based on loan type priorities
## 🔧 Technical Implementation
### Core Algorithms
**Suitability Score Calculation:**
```
S(P,B) = wE·E(P,B) + wI·I(P,B) + wP·P(P,B) + wA·A(P,B)
```
Where:
- E(P,B): Eligibility component (0-1)
- I(P,B): Interest rate favorability (0-1)
- P(P,B): Processing efficiency (0-1)
- A(P,B): Affordability component (0-1)
- w: Loan-type specific weights
**Financial Ratios:**
- Debt-to-Income Ratio (DTI): `CurrentEMI / MonthlyIncome`
- Loan-to-Income Ratio (LTI): `LoanAmount / AnnualIncome`
### Dependencies
The main notebook requires the following Python libraries:
```
scikit-learn>=1.3.0
xgboost>=1.7.0
pandas>=2.0.0
numpy>=1.24.0
matplotlib>=3.7.0
seaborn>=0.12.0
plotly>=5.14.0
jupyter>=1.0.0
```
Install all dependencies with:
```bash
pip install scikit-learn xgboost pandas numpy matplotlib seaborn plotly jupyter
```
## 📊 Usage Examples
### Example 1: High-Income Professional
Open the notebook and locate the applicant profile section. Modify as needed:
```python
# In the notebook, find and modify the profile dictionary
profile = {
'name': 'Arjun Sharma',
'age': 42,
'monthly_income': 150000,
'cibil_score': 810,
'loan_type': 'Home',
'loan_amount': 8000000,
'employment_years': 15,
'existing_loans': 0
}
```
### Example 2: First-Time Car Buyer
```python
# Modify the profile in the notebook for car loan scenario
profile = {
'age': 28,
'monthly_income': 45000,
'cibil_score': 720,
'loan_type': 'Car',
'loan_amount': 800000,
'employment_years': 3
}
```
**Note**: All examples and modifications should be made directly in the Jupyter notebook.
## 🎨 Visualization Features
The system provides comprehensive visualizations:
- **Eligibility Radar Charts**: Multi-bank eligibility probability comparison
- **Suitability Score Bar Charts**: Detailed scoring breakdown
- **Interest Rate Comparisons**: Side-by-side rate analysis
- **Processing Time Charts**: Bank efficiency comparisons
- **Recommendation Explanations**: Natural language justifications
## 🔬 Research Contributions
This work contributes to the field through:
1. **Novel Dual-Model Classification**: Strategic combination of XGBoost and Random Forest
2. **Multi-Criteria Suitability Framework**: Holistic evaluation beyond basic eligibility
3. **Dynamic Weighting System**: Loan-type specific prioritization
4. **Explainable AI Integration**: Transparent recommendation reasoning
5. **Comprehensive Validation**: Real-world scenario testing across diverse user profiles
## 📚 Documentation
- [Research Paper](docs/Smart_Lending_Ally_Research_Paper.pdf) - Complete IEEE conference paper
- [Jupyter Notebook](smart_lending_ally.ipynb) - Full implementation with examples
- Sample Outputs - Available in `docs/sample_outputs/` folder
## 🤝 Contributing
We welcome contributions! To contribute:
1. Fork the repository
2. Create a new branch for your feature
3. Make changes to the Jupyter notebook
4. Test your modifications
5. Submit a pull request with detailed description
### Development Guidelines
- Follow PEP 8 style guidelines for Python code
- Add comments and markdown cells to explain complex sections
- Test with different applicant profiles
- Update documentation as needed
## 📄 License
This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
## 📞 Contact
**Research Team - Jain (Deemed-to-be University)**
- Rahul Rajasekharan Menon - 22btrcn222@jainuniversity.ac.in
- Tejas Rao M N - 22btrcn296@jainuniversity.ac.in
- Nikhil Reddy K - 22btrcn188@jainuniversity.ac.in
- Vishal Kumar P - 22btrcn194@jainuniversity.ac.in
- Nakul Govind S - 22btrcn180@jainuniversity.ac.in
**Faculty Supervisor**
- Dr. A. Balajee - balajee.a@jainuniversity.ac.in
## 🙏 Acknowledgments
- Jain (Deemed-to-be University) for research support
- Reserve Bank of India (RBI) for financial data insights
- National Centre for Financial Education (NCFE) for literacy statistics
- Indian banking sector for providing realistic lending criteria
## 📈 Future Work
- Integration with real-time banking APIs
- Regional loan product variations
- Additional specialized loan categories
- Mobile application development
- Advanced risk assessment models