Predicting Default Risk of Lending Club Loans

Abstract

Lending Club is a peer-to-peer lending company, the largest of its kind in the world with $11.1 billion originated loans. It is an online lending platform where borrowers are able to obtain loans and investors can purchase notes backed by payments based on loans. In this paper, we will attempt to predict the expected returns for loans to a given borrower. Since a loan default will result in a loss of both principal and interest, we will attempt to maximize our returns by predicting the probability of default of the borrower so as to help avoid investment in those high-risk notes. he 2-year U.S. Treasury bond yield hovers below 1%, so Lending Club offers an attractive alternative to bonds for steady investment income. Lending Club offers loans of various grades they assign that correspond to specific interest rates for investors. The higher the interest rate, the riskier the grade. The risk comes in the form of defaults whenever a loan defaults, investors end up losing a portion of their investment. We believe that there is inherent variation between loans in a grade, and that we can use machine learning techniques to determine and avoid loans that are predicted to default. The Lending Club dataset contains a comprehensive list of features that we can employ to train our model for prediction. The dataset includes detailed information for every loan issued by Lending Club from 2007 to 2015, including a borrower’s annual incomes, zip codes, revolving balances, and purpose for borrowing. We will train and test a range of models in an attempt to identify the best performing model.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 14 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.