| We often encounter various types of decision-making,corresponding to different decision-making rules.However,in the risk-based decision-making,the decision-maker often encounters the problem that it is difficult to accurately estimate the occurrence probability of each natural state.Therefore,this paper constructs a machine learning algorithm driven intelligent risk decision-making method,and studies the risk decisionmaking problem in P2 P network loans as an example.The main research work of this paper includes the following three parts:First,the traditional decision tree is used to analyze the risk decision-making of P2 P investors,and the idea of using GBDT series algorithm in machine learning to predict the borrower’s default risk is proposed to drive the risk decision-making.Second,based on the 2018 data of lending Club platform,the default risk of borrowers is measured by using the GBDT series algorithm.In view of the growing data dimension in the era of big data,the GBDT series algorithm of coupling factor analysis is proposed,and its measurement results are compared with the original model in multiple directions.The results of vertical comparison between the original model and the coupled factor analysis model show that the accuracy,harmonic mean value and operation time of the latter are better than the former,so the latter can better measure the borrower’s default risk.The results of the cross comparison of the coupled factor analysis model show that the accuracy(99.92%)and harmonic mean(99.95%)of FA-XGBoost are slightly better,but the running time(8.2839 seconds)of FA-LightGBM is far less than that of FA-GBDT and FA-XGBoost.Therefore,when measuring the borrower’s default risk,FA-LightGBM is recommended when the data size is too large,and FA-XGBoost is recommended in other cases.Thirdly,combined with the conclusion of the second step,FA-XGBoost is used to drive the risk decision-making of investors in P2 P.The results of empirical analysis show that the accuracy of decision-making made by P2P investors based on FA-XGBoost model is 99.92%.Therefore,the risk decision-making method driven by machine learning algorithm proposed in this paper can effectively help investors to make more favorable decisions and successfully realize the intelligent risk decision-making. |
