Research Of Financial Distress Prediction Based On Factor Analysis And Support Vector Machine

Related statistic analysis method and support vector machines (SVM) areemployed in this paper to build a comprehensive and optimalizing SVM model forfinancial distress prediction without restrictions in scale and industry. Comparing withother popular financial distress prediction model in the same experiment condition, wecan observe that the optimal SVM model is feasible and effective. Meanwhile, in orderto achieve the purpose of servicing on the enterprise managers’ and other stakeholders’decision-making, we have integrated three models into the financial distress predictionsystem which plays the role of financial prediction.The concrete content of this research is organized as follows: The researchsituation of financial distress prediction and enterprise performance evaluation internaland oversea is described in detail from two aspects of indexes and methods, and thereasons and performances of enterprise financial distress are analyzed. The principles offinancial distress prediction indexes selecting method based on cash flow are putforward. The statistical learning theory and support vector machine theory is explained,and the financial statement data from A-share companies listed on Shanghai andShenzhen stock exchange is employed to extract the principal component based onfactor analysis. A series support vector machines financial distress prediction modelbased on RBF kernel function, Sigmond kernel function, polynomial kernel function,the linear kernel function and different kernel parameter, are proposed to predict thecorporate financial distress. We compared the financial distress prediction results at thesame experiment condition with BP, logistic model in order to verify the experimentresults.The conventional classic statistic regression mothod and artificial neural networkmethod were produced into the financial distress predicting field under the samecondition in experiment in order to identify the effectiveness of the proposed hybridfactor analysis and support vector machines model. According to the analysis of thedemonstration, the forecast precision of warning model of financial crisis based onfactor analysis and support vector machines is higher than other existed model, such asBP neural network and Logistic regression model.