UPI Transaction Amount Recommendation System

📋 Project Overview

A comprehensive machine learning system that recommends optimal transaction amounts for UPI payments based on user behavior patterns, transaction categories, and contextual factors.

🎯 Key Features

• Personalized Recommendations: Amount suggestions based on individual user spending patterns
• Context-Aware: Considers time, location, category, and payment method
• User Segmentation: Clusters users into spending behavior groups
• Real-time Predictions: Fast inference suitable for production environments
• New User Handling: Category-based recommendations for users without history

📊 Dataset

• Size: 10,000 synthetic UPI transactions
• Users: 1,000 unique users
• Categories: 10 transaction types (Food, Transportation, Shopping, etc.)
• Time Period: Full year of transaction data
• Features: 15+ engineered features for ML models

🤖 Machine Learning Models

Model	MAE (₹)	RMSE (₹)	R² Score
Random Forest	475.68	776.90	0.595
Gradient Boosting	478.66	761.18	0.611
Linear Regression	786.49	1142.29	0.124

🚀 Quick Start

Installation

# Clone the repository
git clone [repository-url]
cd upi-recommendation-system

# Install dependencies
pip install -r requirements.txt

Usage

from upi_recommendation_system import UPIRecommendationSystem

# Initialize the system
recommender = UPIRecommendationSystem()
recommender.load_data('upi_transactions_dataset.csv', 'user_behavior_profiles.csv')

# Get recommendation
recommendation = recommender.recommend_amount(
    user_id='USER_0001',
    category='Food & Dining',
    location='Mumbai',
    payment_method='PhonePe',
    hour=13
)

print(f"Recommended Amount: ₹{recommendation['recommended_amount']}")

Run Demo

python upi_recommendation_system.py

📁 Project Structure

upi-recommendation-system/
├── upi_recommendation_system.py      # Main application
├── upi_transactions_dataset.csv      # Transaction data
├── user_behavior_profiles.csv        # User analysis
├── feature_importance_analysis.csv   # ML insights
├── project_info.json                # Project metadata
├── api_examples.json                 # API samples
├── requirements.txt                  # Dependencies
└── README.md                         # Documentation

🎯 User Clusters

Cluster	Description	Avg Amount	Users
0	Conservative Spenders	₹829	162
1	High-Value Users	₹1,692	146
2	Frequent Small Transactions	₹657	227
3	Active Users	₹986	267
4	Balanced Spenders	₹1,208	198

🔍 Feature Importance

Top factors influencing recommendations:

1. Transaction Category (53.6%)
2. Amount Variability (15.8%)
3. Day of Month (3.7%)
4. User History Length (3.5%)
5. Median Amount (3.3%)

📈 API Endpoints

Recommendation

POST /recommend
{
  "user_id": "USER_0001",
  "category": "Food & Dining",
  "location": "Mumbai",
  "payment_method": "PhonePe",
  "hour": 13
}

User Insights

GET /user-insights/{user_id}

🛠️ Technical Details

• Algorithm: Random Forest Regression
• Features: 15 engineered features
• Performance: 59.5% variance explained
• Speed: Sub-second inference
• Scalability: Handles 1000+ users efficiently

📊 Business Impact

• Improved UX: Reduces amount entry friction
• Increased Engagement: Personalized experience
• Fraud Prevention: Anomaly detection capabilities
• Data Insights: User behavior analytics

🔮 Future Enhancements

• Deep learning models for complex patterns
• Real-time model updates
• A/B testing framework
• Mobile app integration
• Advanced fraud detection

📝 License

This project is licensed under the MIT License.

---

Built with ❤️ for the fintech community