Dynamic Financial Distress Prediction Modeling Based on Slip Time Window and Multiple Classifiers

From a new view of financial distress concept drift, this paper attempts to put forward a new method for dynamic financial distress prediction modeling based on slip time window and multiple support vector machines (SVMs). A new algorithm is designed to dynamically select the proper time window to handle concept drift, and then a dynamic classifier selection method is used to build a combined model. With totally 642 samples from Chinese listed companies, which include ST companies from 2001 to 2008 and their paired non-ST companies, the empirical study is carried out by simulating the process of time passage. The results indicate that slip time window and multiple SVMs method can effectively adapt the financial distress concept drift. This combined model is significantly better than the single model build on the adaptive time window, and they are both better than static models.