| Summary: | 碩士 === 東吳大學 === 經濟學系 === 102 === Recently, because of being opener between Taiwan and China and internationalization, more and more domestic or foreign companies choose Taiwan to build factories and invest money, foreign capital continues inflowing, for this reason, Taiwan has become one of important investment market in Asia. With more and more intense competition in market, how business can survive and develop from the variable economic environment is a crucial issue. Also, because fast change of overall economy environment, it increases the possibilities of financial crisis to business year after year. Hence, building a Financial Distress Early Warning Model efficiently is taken seriously on academics and business community.
It is always traceable when enterprise occurs financial crisis, the reasons a will be reflected on financial statement of the enterprise. Therefore, the research will continue using predictable financial ratio as input variable from documents domestically and exotically, as well as adopt traditional and artificial intelligence both, to build Prediction Models for Financial Distress, hoping it will select out the one with best effect and to provide relative units or people for warning before crisis happening and becoming more serious.
Therefore, this study in addition to Back Propagation Neural Network (BPN) outside and using Fruit Fly Optimization Algorithm combines general repression neural network (FOAGRNN), and Support Vector. Regression (FOASVR), as well as with traditional measure method - logistic regression(Logit) to build Prediction Models for Financial Distress.
Displayed by the result of the research: FOASVR model performs best obviously, the next ones are FOAGRNN and BPN, the last is Logit,. the ways adopting artificial intelligence are all better than traditional measure methods, it also testifies the advantage of AI.
Key words: Fruit Fly Optimization, General Regression Neural Network, Logistics Regression, Financial Distress Early Warning Model, ROC Curves.
|
|---|
