A Comparison Research of the Traditional Econometric Method and the Genetic Fuzzy Neural Networks -- the Securities Business Credit of Taiwan as an Example

• Main Authors: Yi-Fang Ko, 柯懿芳
• Other Authors: Wei-Yuan Lin
• Format: Others
• Language: zh-TW
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/34161891361794369371

碩士 === 東吳大學 === 經濟學系 === 92 === As new financial markets are constantly evolving under the government policy of internationalization and globalization, Taiwan has become the main role of the stock markets in the Asian. But in the economic development of Taiwan, it has hidden some crisis for a long time. Though it will not immediately result into bad effects in short-term, but if confront the economic depression, the possible bankruptcy of the industries could happen, and lead chain reaction of all monetary system to collapse. So, to forecast the credit rating of securities and companies, to warn the related organizations in advance, and to find the ways to prevent the crisis occur beforehand are especially important. This paper uses three different methods of Artificial Intelligence (AI) to build the framework. We hope by these credit rating models could provide alternatives for the financial crisis alarm system. The research is based on the securities business character, and used the principle of Genetic Algorithm as Darwin evolution to build the basis of the networks framework. Besides, the fuzzy logic and Back Propagation Networks are also applied to learn and evaluate this system. We hope these improvements will promote the model forecast ability. The traditional Econometric method is good at dealing with structure, but the environment of financial markets is so complicated and change dramatically. Therefore, its data do not conform to the assumptions of traditional Logit Regression model. However, the limitations of the traditional method used in financial crisis alarm system, but we can’t ignore its ability of sorting and explaining the facts. In the empirical study, we detected the potential problems of Taiwan securities business credit rating, both traditional and AI. We also show the superiority of AI. For the forecast ability, the AI is obviously better than traditional Logit Regression model, and the Genetic Fuzzy Neural Network is the best one in all AI models.