Improved SVM-KNN Model On Credit Risk Assessment

With the rapid development of the social economy,more and more people choose to obtain benefits in the form of investment.In many cases,the lack of investment funds has led creditors to choose bank loans,and the credit evaluation of lenders has become an important part of commercial bank credit.Finding a reasonable and efficient credit evaluation model has important practical significance for promoting the construction of China's credit system and the development of socialist market economy.Based on the credit risk assessment data,this paper discusses the classification effect of the support vector machine(SVM)and SVM-KNN combination model,and uses the grid search algorithm and particle swarm optimization algorithm to optimize the parameters of the combined model.The combination model after parameter optimization has a good prediction effect in the empirical analysisThe main research contents of this paper include:(1)Introduce the source of the data,and descriptively analyze the data.According to the different dimensions of the data attributes,the data is normalized to improve the accuracy of the model,and the character data is normalized after assignment.(2)Apply SVM classification algorithm for data classification,the empirical analysis and comparison were carried out under different training sample sizes and different kernel functions.The effects of different training sample sizes and different kernel functions on the classification accuracy of the model were analyzed.(3)The nearest neighbor(KNN)classification is performed on the data,and the KNN classifier obtains a good classification effect on the test set.Because the SVM classifier is easy to classify the sample points that are close to the classification flat,the KNN classification is performed on the points where the distance classification flat is relatively close,that is,the combination model of SVM-KNN is used to classify the data,and the empirical analysis and comparison are performed.The accuracy of the forecast results is higher than before(4)Because the regularization parameters and kernel function parameters of SVM-KNN classification algorithm have a great influence on the classification accuracy of data,this paper proposes an improved SVM-KNN model,which uses grid search algorithm and particle swarm algorithm to optimize parameters of the combined model,and two sets of optimal parameters are obtained respectively.Through the empirical analysis and comparison,the improved SVM-KNN model can effectively improve the classification accuracy of the dataset.