🧠 AI BNPL Credit Risk Engine

An AI-powered credit risk assessment system for Buy Now Pay Later (BNPL) platforms that predicts default risk, assigns credit scores, and dynamically adjusts loan limits.

---

🚀 Overview

BNPL platforms provide instant credit at checkout, but this introduces significant default risk due to limited credit history and lack of proper risk assessment.

This project builds a machine learning-based credit risk system to:

• Predict loan default probability
• Assign a credit score (0–100)
• Categorize users into risk bands
• Dynamically determine loan limits

---

📊 Dataset

• Total labeled loans: 29,692
• Training set: 23,753
• Test set: 5,939
• Average loan amount: $619.68
• Average APR: 5.28%
• External bureau coverage: 82.6%

---

🧠 Features Used

• clearfraudscore – External credit/fraud score
• repayment_ratio – Debt burden ratio
• nPaidOff – Number of loans repaid
• apr – Interest rate
• leadType_risk – Acquisition channel risk
• loanAmount – Loan size
• payFrequency – Repayment frequency
• ind_totalnumberoffraudindicators – Fraud signals

---

🤖 Model

• Model Used: Random Forest (scikit-learn)
• Why: Best performance among tested models

📈 Performance

Metric	Value
ROC-AUC	0.7273
Recall (Bad Loans)	69%
Precision (Bad Loans)	74%
Accuracy	68%

---

💳 Credit Scoring System

Formula: Credit Score = (1 - Default Probability) × 100

Risk Bands:

• Healthy: 60–100 (Low Risk)
• Cautious: 40–59 (Medium Risk)
• Risky: 0–39 (High Risk)

---

💰 Loan Limit Engine

Risk Level	Loan Limit
Healthy	$1,000 – $3,000
Cautious	$300 – $1,000
Risky	$0 (Auto Reject)

Adjustment Rule:

• +5% per successful repayment
• Max +25%

---

📉 Business Impact

• Reduces credit losses by identifying high-risk borrowers early
• Improves approval quality and consistency
• Enables scalable and automated underwriting
• Supports personalized credit allocation

---

🏗️ System Architecture

• Frontend: Node.js
• Backend: Django REST API
• Model: scikit-learn (Random Forest)
• Database: MySQL (Amazon RDS)
• Deployment: AWS + Alibaba Cloud

---

🧪 How to Run

# Clone the repo
git clone https://github.com/your-username/your-repo.git

# Navigate into project
cd your-repo

# Install dependencies
pip install -r requirements.txt

# Run app
streamlit run app.py#