Research On Credit Risk Early Warning Of P2P Network Lending Platform Based On Machine Learning

The P2P online lending industry has broken through the limitations of time and space in the traditional lending industry with the underlying technology of the Internet.From the perspective of the global situation,it is a useful supplement to the financial system.However,the continuous phenomenon of "thunderstorms" in the domestic P2P online lending industry reveals this There are still many risks in the industry,among which the credit risk of industry entities(P2P platforms)is particularly prominent.Therefore,combining the characteristics of the times and academic methods to explore how to realize the early warning of the credit risk of the P2P network lending platform has certain academic value and practical significance.This paper uses machine learning technology to study the credit risk of P2P online lending platforms based on webpage data captured by crawler technology.First,it discusses the related theories of machine learning and P2P platform credit risk,laying a theoretical foundation for later data research;Secondly,combing the development history and status quo of domestic and foreign industries through relevant data collection,document integration,etc.,laid the realistic foundation of this article's research;Then,write code in R language to perform data preprocessing,analysis and feature engineering to realize the preliminary understanding of the data set and the research and exploration of platform credit risk;Then,build BP neural network,support vector machine and convolutional neural network machine learning early warning models respectively,and conduct model training and testing on platform data;Finally,through comparative analysis of the early warning effects of various models,relevant evaluations and policy recommendations are made.At the same time,the project investment period,credit transfer and credit enhancement methods can be seen to a certain extent as measures to alleviate the demand for rigid redemption of Chinese people.The more effectively the platform can cope with the impact caused by rigid redemption,the smaller the corresponding credit risk.Through exploratory analysis,the law between multiple indicators and P2P platform credit risk is obtained.The platform background,registered capital and paid-in capital are all important indicators to show the strength of the platform.Research shows that the stronger the platform,the lower the credit risk.In addition,the credit risk of the platform is closely related to the platform's rate of return and the type of business.Studies have shown that within a small reasonable range,the platform's credit risk will decrease as the platform's return rate increases,but when it exceeds the reasonable threshold of about 10%,the platform's high return is more likely to mean a high-risk area;The credit risk of the platform decreases as the life details of the business type deepen.The encapsulation and dimensionality reduction method of Boruta algorithm not only has the characteristics of simplicity and visualization,but the selected features are relatively objective,which reduces the potential adverse effects of subjective selection features on the early warning effect of the model,and is used in the application of the later machine learning early warning model.An important foundation is laid for optimal performance.At the same time,21 indicators such as average interest rate,bank depository,and ICP certification selected by the algorithm can be focused on in the development of the industry,and coordinated management in many aspects can be realized to help the P2P network lending industry more adapt to the development needs of the international financial environment.Both deep machine learning models and shallow machine learning models can realize early warning of platform credit risks.Among them,the Convolutional Neural Networks(CNN)in deep machine learning performs better overall,and its model performance has the best stability.And the construction and application of platform credit risk early warning models based on machine learning can effectively promote credit risk control in the P2P network lending industry.