Credit Risk Prediction System

Project Overview

This project predicts whether a loan applicant is likely to default on a loan using machine learning. The solution is built as an end-to-end ML system using PostgreSQL, XGBoost, and FastAPI.

#Architecture Diagram

Business Problem

Financial institutions face significant losses when customers fail to repay loans. The objective of this project is to predict customer default risk and classify applicants as:

• High Risk
• Low Risk This helps lenders make better credit approval decisions.

---

Dataset

Home Credit Default Risk Dataset

Source: https://www.kaggle.com/competitions/home-credit-default-risk

Tech Stack

• Python
• PostgreSQL
• Pandas
• NumPy
• Scikit-Learn
• XGBoost
• FastAPI
• Swagger UI
• Git
• GitHub

---

Project Architecture

CSV Files ↓ PostgreSQL Database ↓ Feature Engineering ↓ Feature Store ↓ Data Preprocessing ↓ XGBoost Model ↓ Model Serialization ↓ FastAPI API ↓ Real-Time Predictions

Features Implemented

Data Engineering

• Loaded 11 source files into PostgreSQL
• Created feature engineering pipeline
• Built customer-level feature store

Machine Learning

• Missing value handling
• Label Encoding
• Train/Test Split
• Logistic Regression
• Random Forest
• XGBoost

Model Evaluation

Metrics used:

• Accuracy
• Precision
• Recall
• F1 Score
• ROC-AUC

Deployment

• FastAPI REST API
• Swagger Documentation
• Real-time customer risk prediction

---

Sample API Response

{ "customer_id": 100002, "default_probability": 0.88, "prediction": "High Risk" }

Future Improvements

• Dockerization
• Hyperparameter Tuning
• Model Monitoring
• CI/CD Pipeline
• Cloud Deployment

---

Author

Sanjay Sekar

Data Scientist