A Meta-Learning Approach forBankruptcy Prediction

• Main Authors: Yu-Feng Hsu, 許育峯
• Other Authors: Chih-Fong Tsai
• Format: Others
• Language: en_US
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/71904634317607974931

碩士 === 國立中正大學 === 會計與資訊科技研究所 === 96 === In recent years, information globalization makes a great change on the business operation environment, and the global economic declines which enhance the probability of financial distress. Business bankruptcy will significant affect the economical frame of a country and make huge damage. The purpose of this thesis is to build up a meta-learning classifier by data mining and machine learning techniques, and to compare with single classifiers and stacked generalization in order to establish a more effective bankruptcy prediction model. In particular, the first level classifiers aims at filtering out unrepresentative training data (or outliers) for the second level (meta) classifier. In this thesis, three classification (or supervised learning) techniques are considered, which are neural network (MLP), decision tree (CART) and logistic regression (LR). These three classifiers are used to build up single baseline classifiers, the stacked generalization classifier, and the meta-learning classifier respectively. The comparative results of the Baseline, Stacked generalization and Meta-learning approaches show that the Meta-learning approach perform the best in terms of prediction accuracy as well as the Type I and II errors.