Business Crisis Prediction Based on SVM Multiple Classifiers

• Main Authors: En-Chia Chen, 陳恩加
• Other Authors: Deron Liang
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/86064057805671151394

碩士 === 國立臺灣海洋大學 === 資訊工程學系 === 98 === Recent year, financial variables are widely used to establish diagnosis models for business crises. The combination of feature set and the usage of classifiers in business crisis prediction are widely and continually studied topic in the field of corporate finance. The aim of this research is twofold. First, we expand the range of predictors by adding both popular financial features the prior scholars and features of TEJ (Taiwan Economic Journal) database. We apply data mining techniques to identify five useful predictors, [Taiwan Corporate Credit Risk Index] [continuous four quarterly EPS] [Net Income%-after tax][Pre-Tax Income %] [Debt/ Equity %] which are previously unaware to the community. For concept verification, we compare several scholars’ model with our proposed SVM model. Our experiment indicates that the SVM model based on our proposed feature set outperforms those models based on the recommended feature sets by prior scholars in terms of the prediction accuracy. Secondly, we design a SVM-based multiple classifiers with multiple type of business crises. By combining multiple results of prediction models, the prediction accuracy and flexibility of are better than traditional single classifier.