Prediction of Corporate Financial Distress Using Support Vector Machines

• Main Authors: Chi-Ming Tang, 唐其民
• Other Authors: Chien-Kuo Li
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/78467140027158575149

碩士 === 實踐大學 === 資訊科技與管理學系碩士班 === 96 === The prediction of corporate financial distress has long been an important and challenging research topic. A well-developed prediction system can prevent numerous investors from big loss due to investment in risky company. Banks need to predict the possibility of default of a potential counterparty before they approve a loan. Therefore, banks can also benefit from bankruptcy prediction modeling. In recent years, there are a considerable number of studies on the modeling of corporate financial distress prediction. The prediction of corporate financial distress is a classification problem. There are two most important factors which will affect the performance of the prediction model. The first factor is the research method and the second one is the variables to be put in as the financial indicators. The support vector machine has been widely used in various kinds of classification problems. In this research, support vector machine was used to develop the prediction model of the corporate financial distress, with a view to test the feasibility of the research method. Research results were compared with those from back-propagation neural networks. Most researches on the prediction of corporate financial distress only use financial indicators as the input variable. However, after the 1997 Asian financial crisis, many research results showed that corporate governance was a very important factor leading to the financial crisis. Therefore, both financial indicators and corporate governance factors were taken into account in the establishment of the model of corporate financial distress prediction. After a thorough paper review, 47 financial indicators as well as three non-financial indicators were considered in this research. These 50 indicators were divided into eight categories. Computer simulations were conducted to find the best input variable combination which would lead to the best classification results. Computer simulations showed promising results for the application of support vector machine in the modeling of corporate financial distress prediction. The values of F-measure achieved over 85% for corporate data obtained one year, two years, and even three years prior to the corporate bankruptcy. Research results also showed that better prediction model can be obtained by having both financial indicators and non-financial indicators as the input variables. Investors and business operators can use the prediction model when making investment decisions. Banks can also benefit from this prediction model when making lending decisions.