Research On Companies' Financial Distress Prediction Based On Feature Selection And Support Vector Machine

In the intense competitive market economy, the construction of proper prediction model is of great significance to not only protect and guarantee the benefits for investors and creditors, keep away financial crisis for proprietors, but also monitor the quality of listed companies and risks in stock-market for official administration sections since it will accurately predict business financial distress.Statistical learning theory is a theory of machine learning law dealing with small samples, and it takes into account the requirement of the generalization ability and the most excellent answer in limited conditions. Based on Statistical Learning Theory, a new machine learning method-support vector machine is put forward, and there are some virtues in dealing with the problem of pattern recognition, such as the problems of small samples, high dimensionality, non linearity.This paper gives an integrative introduction to correlative theories and models of financial distress prediction in and abroad. On a basis of the domestic and abroad two main different types of data sets of business financial predicament, the variables analyzed by way of using the statistical identifying method. In this paper using the effective data mining algorithm ~ SVM classifier as a modeling method, and introduce Independent Component Analysis as a feature selection tool to effective select the better ratios of financial indicators from financial data for the establishment of enterprises' financial distress prediction, thereby optimizing and improving the classification model performance based on support vector machines. By domestic and foreign enterprises for the financial distress or bankruptcy of the data empirical analysis and comparison with other methods results, confirmed the validity and practicality of the enterprise financial distress prediction model through independent component analysis and support vector machine classifier established.