none
| Autor: | Chia-Chi Yeh, 葉嘉琪 |
|---|---|
| Rok vydání: | 2011 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 99 In recent years, the global economic recession, followed by failure of business due to poor management, resulting in a domino effect occurred throughout the financial system, with its wide-reaching impacts. In order to avoid loss of the expansion and early attempts to prevent the crisis occurred, the problems of business failure should really be explored. To establish a complete set of financial distress prediction models, companies will be able to predict in advance the possibility of a crisis, and thus make the relevant personnel in some warning before the crisis, to cope with the strategies to prevent the crisis To improve the efficiency of decision-making and analysis, this study adopted the method of data mining. Firstly, data collected before making a pre-processing (data reduction). Stepwise regression, genetic algorithms and self-organizing feature map network is used to reduce the Taiwan listed company's financial variables, governance variable, external rating variable, accountant variable and economic variables (96 variables) to 4 variables. The sample period is from 1995 to 2009, a total of 399 companies are used in this study. Secondly,we apply Eviews, Pythia, and Matlab programs to a variety of methods of artificial intelligence --genetic algorithm, backpropagation neural network, genetic backpropagation network and genetic programming to construct a variety of financial distress prediction models, and compare their performances with the traditional econometric methods --OLS, Logit, Probit, Extreme Value models; Then, through the ROC curve the model prediction accuracy between the sensitivity and specificity of the corresponding relations are explored; Lastly, we conducted the Wilcoxon signed rank test to validate the model prediction performance between their differences. The empirical results showed that use of these forecasting models really can effectively reduce the forecast error rate, and artificial intelligence learning models indeed is better than the traditional econometric models for their predictive ability. Key words: Data Mining, Logit Regression, Stepwise Regression, Self-Organizing Map, Genetic Backpropagation Neural Network, Genetic Programming, Financial Distress Prediction Model, ROC Curves, Wilcoxon Test. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|