Investment Return and Default Prediction in Online Peer-to-Peer Lending

Capstone Project 1 (Springboard - Data Science Career Track)

Chiyuan Cheng (05/2020)

Summary

• Peer-to-peer (P2P) lending is the new practice of lending money to individuals or small businesses via online service that matches lenders with borrowers.
• Lending Club (LC) is the world’s largest P2P lender according to their issued loan volume and revenue. However, in contract with the traditional investment, P2P lending presents a higher credit risk, because the borrower has a higher chance to not pay off his/her loan, leading to the loan default. This motivates us to build machine-learning models to predict the credit risk and optimal investment return with the LC historical loan dataset.
• Random Forest and Gradient Boosting models perform the best with default prediction and investment return prediction, while Gradient Boosting performs slightly better than Random Forest.
• The best investment strategy achieves 40% Annualized Return of Investment, which is 8X lending club benchmark.
• P2P lenders can take advantage of the predictive models to help investors to make smart decisions when evaluating loan application.

Report

• Presentation Slide
• Final Report

Jupyter Notebook

1. Data Access
2. Data Processing 1
3. Data Processing 2
4. Exploratry Data Analysis
5. Statistical Analysis
6. Machine Learning (Classication)
7. Machine Learning (Regression)