Financial crisis prediction based on distance to default and feature weighted support vector machine

Abstract

In order to keep the market run regularly, the prediction of financial crisis becomes necessary and urgent. A new risk rating method based on distance to default (DD) and order statistics (OS) is established to classify listed companies into three ratings according to their financial risks. In addition, financial indicators are weighted based on DD and grey relational degree. On the basis of the new method, financial crisis prediction is researched based on feature weighting SVM (DD-FWSVM) model with three classifications in the study. The experimental analysis is conducted based on the listed companies in the Growth Enterprises Market (GEM) of China at last and the result demonstrates that our model has better performance in financial crisis prediction when compared with other methods.