Risk Assessment Model Study Of P2P Lending Platform Based On Stacking Integration Strategy

With the rapid development of P2P online lending platforms,incidents such as platform running,economic investigation and intervention and withdrawal difficulties also occur frequently.That eauses losses for investors and brings a bad impact for the P2P lending industry.Assessing the risks of PZP lending platforms is a major problem faced by investors,regulators and platform managers.However,in the previous research,decision tree algorithm and other single algorithm are often used for model construction,with low prediction accuracy and poor applicability.Therefore,it is important for us to study the risk assessment model of P2P online lending platforms.This paper takes the risk of P2P online lending platforms as research object.First,we constructed P2P platform risk evaluation index based on P2P oiline lending platform lending risk analysis,use Scrapy crawler to crawl the data platform and establish a complete set of processes from the data up to data cleaning and data analysis.And we use k-means clustering algorithm to divide platform's risk level.Then we establish and test the risk assessment model of P2P online lending platform based on Stacking integration strategy.Lastly,suggestions are made to investors,regulators and platform managers respectively based on the research results.(1)A new risk assessment index system for P2P online lending platforms is established.In view of the existing risk assessment index system of index selection is imperfect,index data acquisition is difficult,and subjectivity strong risk hierarchies,we analysis the credit risk,legal risk,operation risk,liquidity risk and market risk to explore the influence factors of risk behind.Scrapy framework is used to crawl data,and we conduct correlation analysis and descriptive statistical analysis on the cleaned data;Then,we use k-means clustering algorithm to classify the platforms based on the risk indicators.According to the clustering results,the platforms were divided into five categories:safe platform,low-risk platform,medium-risk platform,high-risk platform and heavy risk platform.(2)The risk assessment model of P2P online lending platform based on Stacking integration strategy is constructed.On the decision tree algorithm is easy to deal with discrete variable but is susceptible to noise interference produced data overfitting phenomenon and the support vector machine(SVM)algorithm is susceptible to noise interference data but it hard to choose kermel function,using Stacking integration strategy,to the decision tree algorithm and support vector machine(SVM)algorithm as the base classifier,support vector machine(SVM)algorithm as meta classifier,constructed the risk assessment model of P2P lending platform based on Stacking integration strategy,and the validity of the model is verified by comparing to the model based on decision tree algorithm and the model based on support vector machine algorithm model on prediction accuracy,ROC and AUC comparative three indexes.In addition,the correlation between the risk level of platform and the operation type of platform is analyzed.The results show that normal operation is the main type of platform operation in safe platform,while withdrawal difficulty and running are the main types of platform operation in high risk platform and heavy risk platform respectively(3)Put forward suggestions on how to prevent and control platform risks for investors,regulatory authorities and platform managers.In view of the lack of data support for the suggestions proposed in previous studies on platform risk management,based on the research results,we suggested that investors should not pursue returns blindly and examine the platform in a multi-dimensional way when choosing a platform.And we suggested that the regulatory authorities should improve the access mechanism of the platforms and implement hierarchical supervision of platforms.And we suggested that managers of P2P online lending platforms should strengthen the examination of borrowers,enhance industrial cooperation and improve the risk control mechanism of the platforms themselves.