Research On Financial Risk Early Warning Model Of Listed Companies Based On Fuzzy Support Vector Machine

With the rapid development of Chinese economy,the securities market and economic market are also present a prosperous scene.In order to expand the scale and improve the social influence and competitiveness,enterprises have chosen to go public.However,the vigorous development of economic market must accompany the complex and changeable market environment.Listed companies seize the opportunity but also accompanied by many challenges.In order to ensure the steady development of the enterprise and reduce the possibility of bankruptcy,it is very important to evaluate and control the financial risk at any time during the production and operation of the enterprise.Therefore,how to establish an effective system to evaluate financial risk has become an important issue in enterprise management.This paper selects some high-tech industry listed companies in Shanghai and Shenzhen stock markets as research objects.On the basis of the research on financial risk by domestic and foreign scholars and combining with the characteristics of high-tech industry,25 financial indexes which can effectively evaluate the risk status are preliminarily selected.By using significance test,17 index variables which can distinguish financial risk from non-financial risk are selected.In order to further reduce the data redundancy and improve the prediction accuracy,the KPCA is used to extract the features of these 17 financial indicators.Finally,12 principal components with a cumulative contribution rate of more than 90% are obtained.Based on fuzzy theory and SVM,a FSVM early warning model based on KNN fuzzy membership degree is constructed,and a multi-classification financial risk early-warning model of FSVM is proposed on the basis of two-classification.In order to further improve the prediction accuracy of the model,the penalty parameters and kernel function parameters in the FSVM model are optimized by GA,PSO and SAPSO.Input preprocessed financial data into four warning models in turn.Through the empirical research of four models,the results show that the SAPSO-FSVM model has a good classification effect on financial data.Finally,this paper puts forward the financial risk warning model of SAPSO-FSVM listed companies.The main research contents include:(1)The preprocessing of financial risk early warning samples is studied.According to the subjective consciousness and the experience of previous scholars,there are great differences between the selected financial indicators.Direct input of data into the early warning model will not only increase the running time of the model,but also reduce the prediction accuracy of the model.In this paper,the primary index is screened by significance test,and 17 indexes which can distinguish the financial risk status are selected.In the case of ensuring the integrity of data information,the KPCA feature extraction of these 17 indexes is carried out.Thus the data redundancy is reduced and the experimental data set of the final input model is obtained.(2)Based on SVM combined with fuzzy theory and K-nearest neighbor algorithm,a FSVM early warning model based on KNN fuzzy membership degree is constructed.Considering the influence of penalty parameters and kernel function parameters on the early warning effect of FSVM model,combined with the characteristics of GA,PSO and SAPSO optimization,the optimal selection of FSVM model parameters are carried out respectively.Finally,four financial risk warning models of FSVM,GA-FSVM,PSO-FSVM and SAPSO-FSVM listed companies are established.(3)The pre-processed sample data set is input into four kinds of early warning models for empirical research.The experimental results show that the accuracy of SAPSO-FSVM in predicting financial risk is 79.29%,which is superior to the other three models.It is suitable for high-tech industry listed companies as a reference for financial risk assessment.