Development of a Hybrid Case-Based Reasoning for Bankruptcy Prediction

Abstract

AbstractThis paper aims to develop an integrated model of predicting business failure, using business financial and non-financial factors to diagnose the status of business, thereby providing useful references for business operation. This study applied Rough Set Theory to extract key financial and non-financial factors and Grey Relational Analysis (GRA) as the approach of assigning weights. In addition, Case-Based Reasoning (CBR) are adopted to propose a new hybrid models entitled RG-CBR (combining RST and CBR with GRA) to compare the accuracy rates in predicting failure. After exploring the TEJ (Taiwan Economic Journal) database and conducting various experiments with CBR, RST-CBR and RG-CBR the study finds CBR, RST-CBR and RG-CBR reporting an accuracy rate in predicting business failure of 49.2%, 59.8% and 83.3%respectively. The RG-CBR boasts the highest accuracy rate while also effectively reducing Type I and Type II error rates.Keywordsdecision analysisdata miningbusiness failure predictionrough set theorycase-based reasoning