The Application Of Factor Analysis And Artificial Intelligence In The Financial Early Warning Of Listing Company

Financial crisis will bring about a series of serious consequences, such as the loss of investors, the unemployment of the employees and the difficult to recover the financial credit. How to use statistical methods, machine learning and data mining to provide effective financial early-warning for China’s listing companies has become an urgent and major issue.In this paper, three main financial early warning models were designed based on statistics and artificial intelligent classifier. Firstly, I made variable selection for the first 30 financial ratios through multi-step factor analysis to determine the financial ratio system of listed company and then constructed a factor analysis financial early warning model. Secondly, An artificial intelligent single classifier for financial early warning model based on analysis LSSVM-PSO was brought forward. The empirical results showed that the LSSVM-PSO combined with muti-step analysis had a better performance. Finally, Three multi-classifiers for financial early warning model respectively based on Adaboost ensemble BP-neural network, Adaboost ensemble decision tree, and Adaboost ensemble SVM were built. The experimental results showed that AdaBoost_DT outperformed AdaBoost_BP and AdaBoost_SVM.The innovation of this paper mainly lies in the following three aspects:First, the factor analysis model of financial early warning is constructed by the method of multi-step factor analysis that highlights the objectivity and the advantages of the factor analysis method in the selection of financial indicators.Second, based on the further expansion of artificial intelligent single classifier,multi-step factor analysis method and artificial intelligent single classifier combined effectively, not only retained the original financial information and simplified the model to the greatest extent, but also greatly improve the running speed and warning effect of the artificial intelligent single classifier.Third, the use of BP neural network, decision tree and support vector machine as weak classifiers to ensemble Ada Boost multi-classifier, then analysized and compared the results of the three classifiers, finally selected the best artificial intelligent multi classifier financial early warning model.