Random Forest Prediction Model Of P2P Network Lending Default

In the 21 st century,Internet and computer technology have become increasingly important to human social life.Internet+thinking has also spread to all walks of life in society,and has also brought about tremendous changes in the financial industry.P2 P Internet lending platform has emerged as a result.It can provide a way for people with idle funds to find value-added and increase value.It can provide lending services to companies or individuals in financing trouble.P2P network lending is provided by the internet credit company,and the lenders and the lenders are free to bid and form a form of syndication.Internet credit originated in the United Kingdom.Due to high interest rates and low investment and loan thresholds,it quickly developed into other countries and flourished.By the end of 2005,there were about 2,300 online banking platforms registered in China,and the amount of online banking industry transactions exceeded 10,000 Billion.However,in the rapid development of the P2 P online loan industry,many problems have also been exposed,especially the issue of credit risk.Because the P2 P network lending does not require the participation of third parties,but the borrower and the lender directly conduct transactions,and our country's citizen credit system has not been standardized,and the exist problem of information asymmetry,the phenomena of breach of contracthas happened frequently,which ultimately brings huge losses to the network lending platform and lenders.Therefore,establishing a prediction model for breach of contract is of great significance to the healthy and stable development of the Internet lending industry.In this paper,a random forest algorithm is used to establish the default model with the loan information data of Prosper P2 P network loan platform.Because the original data is responsibleand the quality is difficult to guarantee,At first,the original data is need to be preprocessed,including the standardization of continuous numerical variables,Quantification of categorical variables,deletion and interpolation of samples with missing values.The person who commits a violation is a relatively small part,so the data is unbalanced data,This paper uses the SMOTE algorithm to balance the positive and negative sample proportions.In the modeling process,heuristic method is adopted to select the optimal parameters of the model,the overall performance of the model is evaluated using a Confusion Matrix,and the optimal classification threshold is selected with the ROC curve to improve the performance of the model,It is hope that it can provide decision basis for the P2 P online loan platform.