Machine Learning on Credit Risk for Lending Club Business

Abstract

Under the big data era, human beings attach much importance on machine learning. The purpose of this paper is to identify the way to use different methods of machine learning for data analysis and data mining when the companies in lending industry is faced with credit risk. To be specific, the methods are judged from the indicators (e.g., accuracy score and AUC score), and utilized to recognize which loans are bad loans, so as to facilitate the company to make better decisions. The methods including Random Forest, Gaussian Naive Bayes, and Artificial Neural Network are discussed. According to the analysis, all models have an accuracy around 60-70%, while each show different tendency in classifying results. Further optimization that can be applied in the future studies is suggested in the paper. The overall value and reputation of a company will be improved with good credit risk management. Therefore, a good method of credit risk management (e.g., accurately identifying good loans and bad loans) is classifying and analyzing the existing data of the company via different algorithms, and eventually compare them. These results shed light on guiding further exploration of evaluating the creditworthiness in the lending field.