Developing a business failure prediction model via RST, GRA and CBR

Abstract

The prediction of business failure is an important and challenging issue that has served as the impetus for many academic studies over the past three decades. The widely applied methods to predict the risk of business failure were the classic statistical methods, data mining techniques and machine learning techniques. In addition to diagnosis and classification, Case Based-Reasoning (CBR) is an inductive machine learning method that can be applied to replace statistical models. Concerning the fact that the attribute extraction and weighting approach could enable CBR to retrieve the most similar case correctly and effectively, this paper proposes a Hybrid Failure Prediction (HFP) model by applying Rough Set Theory (RST) and Grey Relational Analysis (GRA) as data preprocessors to strengthen the effectiveness of CBR predicting capability. After exploring the data from TEJ database and comparing it with three models, CBR, RST-CBR, and HFP, the results show that our model is the most accurate and effective model in predicting business failure.