A Case-Based Reasoning Approach to Business Failure Prediction

Abstract

AbstractTremendous efforts are spent and numerous approaches are developed for predicting business failures. However, none of the existing approaches is dominant with respect to the accuracy and reliability of the prediction outcome. Contradictory prediction results are often present when different approaches are used. Also, explanation and justification of a prediction is often neglected. This paper reviews different approaches and presents a framework of a case-based reasoning (CBR) approach to business failure prediction by integrating two techniques, namely nearest neighbor and induction. It is unrealistic to assume that all attributes are equally important in the similarity function of nearest neighbor assessment. To avoid the inconsistency of subjective preferences of human experts, induction is used to find the relevancy of the attributes for nearest neighbor assessment in the case matching process. The approach is expected to provide an accurate prediction with justification, which is useful and beneficial to stakeholders of the companies.KeywordsLinear Discriminant AnalysisInformation GainAttribute WeightLinear Probability ModelBank FailureThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.