A Research On Financial Crisis Prediction Of Listed Companies Based On Particle Swarm Optimization And Support Vector Machine

In this paper, current situation about the research on financial crisis prediction at home and abroad was analyzed in detail. On the basis of previous researches and the actual situation of the enterprises and capital market of China, the research method of this problem was determined. In the previous studies on financial crisis prediction, the traditional statistical model was widely used. But, unfortunately, these models were restricted by the conditions such as strictly assumption, large and strictly constrained samples and so on, when you want to get good result. So, these strictly constrained conditions reduced the ability of Generalization of the model. Further more, these models have many shortages in accuracy and application. As the problems mentioned above and the excellent performance and superiority of Particle Swarm Optimization algorithm (PSO) and Support Vector Machine (SVM) in the field of parameter optimization and classification, in this paper, the approach of PSO combined with SVM was used as the core idea of the research on financial crisis prediction of listed companies of China.In this paper,100 List companies from Shanghai Stock Exchange Market (SSE) and Shenzhen Stock Exchange Market (SZSE) were selected, which accrosses differently fields, as the research sample. On the basis of the sample, the hybrid PSO-SVM model was established, in which LS-SVM was used. And make an analysis on these samples. What's more, in order to validate the accuracy and ability of generalization of the PSO-SVM model, in this paper, different indicator was used to evaluate the model. The research result shows:the model established in this paper reaches the accuracy of 90.5%, by using the data that a year before financial crisis occurs. What's different between the traditional statistical model and PSO-SVM nodel is that the later model combined the advantages of AI technology with the advantages of statistic. So, this model has the both advantages of accuracy and generalization. So the PSO-SVM model has the great value of application.