Early Warning Research On Special Treatment Of Listed Companies Based On Support Vector Machine And R-type Clustering

Firstly,this paper analyzes the reasons why listed companies are specially treated and the status of domestic researches and abroad researches,and introduces the support vector machine learning method that plays a leading role in this paper.Then it constructs the Special Treatment(referred to ST)warning index system of listed companies,which is divided into two rounds of indicators screening:In the first round,the support vector machine method is used to screen out the indicators that can significantly affect the ST status of listed companies.The second round is to eliminate the information redundancy index by R-type clustering method,and then use the ROC curve method to test the effectiveness of the early warning indicator system.A ST early warning indicator system for listed companies that can distinguish between ST companies and non-ST companies without information redundancy is constructed.Secondly,the ST warning model of listed companies is constructed.Based on the support vector machine two classification model,a new model is established.Listed company ST early warning model,and use MATLAB 2015a and SPSS 20 to combine the financial data of the listed company of CSMAR Data Center to apply the constructed model.Finally,through the empirical results of the model,find out the main influencing factors that cause the listed company been ST,and provide listed companies with some suggestions to avoid being ST.The main innovations of this paper are as follows:Firstly,select the Levene variance homogeneity test statistic W value to identify the ST status identification ability of the early warning indexes,and use the R-type clustering method to cluster indexes according to the criteria layer,and retain the indexes with the largest W value in each type of indicators to delete the information redundancy indicators and in the same time,the remaining indicators retain the early warning indicators with significant ST status identification capabilities.Secondly,in the construction of ST early warning indicator system,the existing researches are mainly to select indicators manually,which may result in the loss of some important indicators and the existence of information redundancy indicators,this paper uses the dual screening mechanism of support vector machine method and R type clustering method to construct early warning index system;Thirdly,listed company is ST or non-ST is nonlinearly related to early warning indicators,the existing researches mostly build multivariate early warning model based on linear relationship,which distort the actual relationship and reduce the prediction accuracy of the model,this problem is solved by using the Gaussian radial basis kernel function in the support vector machine.Fourthly,the existing financial early warning models are mostly based on traditional statistical models,such as linear classification analysis,whose calculation is complex,the precision is not high,and the generalization ability is weak,this paper chooses the support vector machine model of artificial intelligence method,which is simple in calculation,high in precision and strong in generalization ability.