🧠 AI Lending Risk Assessment

A step-by-step machine learning project to assess loan default risk using borrower data

📘 Read full project write-up: Building an AI Lending Model – Step-by-Step Guide

This is a web application deployed on Google App Engine that provides real-time loan default risk assessment using a machine learning model deployed on Vertex AI. The application takes loan details as input and returns the probability of default along with risk assessment metrics.

Features

• Real-time loan default risk prediction
• Web-based interface for loan officers
• Integration with Vertex AI for model inference
• Comprehensive input validation
• Detailed risk assessment results including:
  • Default probability
  • Risk level classification
  • Confidence score
  • Request tracking

Prerequisites

• Google Cloud Platform account
• Python 3.9
• Google Cloud SDK
• A trained model deployed on Vertex AI

Project Structure

lendingapp/
├── main.py               # Main application file
├── app.yaml             # App Engine configuration
├── requirements.txt     # Python dependencies
├── artifacts/
│   ├── numeric_features    # List of numeric features
│   ├── selected_features   # List of all features used by model
│   └── scaler.pkl         # Fitted StandardScaler object
└── templates/
    └── index.html         # Frontend interface

Configuration

1. Update app.yaml with your project details:

env_variables:
  PROJECT_ID: "your-project-id"
  LOCATION: "us-central1"
  ENDPOINT_ID: "your-endpoint-id"
  GOOGLE_CLOUD_PROJECT: "your-project-id"

2. Ensure your Vertex AI model endpoint is deployed and accessible.

Input Features

The application accepts the following inputs:

• Original Loan Amount ($): 10,000 - 2,000,000
• Credit Score: 300 - 850
• Original Loan-to-Value Ratio (%): 0 - 100
• Debt-to-Income Ratio (%): 0 - 65
• Original Interest Rate (%): 0 - 25
• State: Valid US state codes

Deployment

1. Clone the repository:

git clone [repository-url]
cd lendingapp

2. Install dependencies locally (for testing):

pip install -r requirements.txt

3. Deploy to App Engine:

gcloud app deploy app.yaml

4. View the deployed application:

gcloud app browse

Understanding Results

The application provides:

• Default Probability: The model's predicted probability of loan default
• Risk Level: Classification as "High" (>50%) or "Low" (≤50%)
• Confidence: The model's confidence in its prediction
• Request ID: Unique identifier for tracking predictions
• Timestamp: When the prediction was made

Error Handling

The application includes:

• Input validation for all fields
• Proper error messages for invalid inputs
• Logging for debugging and monitoring
• Error tracking with request IDs

Monitoring

Monitor the application using Google Cloud Console:

• App Engine dashboard for application metrics
• Cloud Logging for detailed logs
• Vertex AI dashboard for model metrics