Design And Implementation Of Network Loan Violation Forecasting System Based On Adaboost-SVM

With the continuous improvement of the level of computer network security and the rapid development of the Internet,P2 P network lending is continuously eroding the Chinese Internet market,and the development momentum is swift and violent.It is a personal to personal loan model that is independent of the Internet financial institution system.Although the development of P2 P network lending has brought great economic benefits,it also exposed a series of problems.Due to the defect of the platform operation of the P2 P network lending platform,there is a lack of participation of third-party agencies in the entire lending process.The lender's risk assessment is mainly based on personal credit.However,due to the lack of an effective and reasonable supervision system and related regulatory platform in China,coupled with the incomplete credit information collection system of the borrower,the borrower's overdue default behavior occurs frequently.Therefore,starting from the network loan data in this context,data mining algorithms are used to analyze the impact of data characteristics on the results of the loan,so that the extracted features have good reliability and sensitivity in the actual classification prediction.This paper uses Support Vector Machine(SVM)as a classifier.The research shows that not only the feature selection will greatly affect the classification effect of SVM,but also the appropriate SVM penalty parameters and kernel function parameters can improve the classification effect.Particle Swarm Optimization(PSO)is used to optimize the parameters of SVM,and an improved particle swarm optimization algorithm is proposed.Based on the improved particle swarm optimization algorithm,the model parameters of SVM base classifier are optimized,and then the AdaBoost algorithm is used to iteratively integrate the above optimized SVM base classifiers,and a network loan default prediction model based on multi-classifier optimization integration is created.Taking the Lending Clud loan data set as the research object,the network loan default forecasting model was built.The experimental results show that compared with the standard SVM algorithm,PSO-SVM algorithm,decision tree algorithm(CART)and random forest algorithm(Random Forest),The combination of AdaBoost and SVM effectively improves the classification accuracy and generalization rate of the entire sample.The accuracy of network loan default prediction using this model is significantly better than other models.Finally,based on this model,this paper designs and implements a network loan default prediction system based on AdaBoost and SVM.After being verified by an independent test set,the system can accurately predict and achieve an accuracy of 92.5%.Both functional and performance can meet system requirements and achieve the desired research results.