A psychometric method for structuring expert knowledge: Application to developing credit analysis expert systems for small–medium companies using nonfinancial statement information

Translating expert knowledge into production rules has been the most difficult and time-consuming feature when building expert systems. In particular, building hierarchical structures, i.e. developing sequential or dominant relationships among production rules is one of the most important and difficult processes. Hierarchical relationships among rules have typically been determined by interviewing human experts. Since this interviewing procedure is rather subjective, however, hierarchically structured rules produced as a result of interviewing are widely exposed to criticism about their validity. We thus need an objective method to effectively translate human expert knowledge into structured rules. For such a method, this paper suggests the order analysis technique that has been studied in psychometrics. In this paper, we briefly introduce the order analysis and explain how it can be applied to building a hierarchical structure of production rules. We also illustrate how bankruptcy prediction rules of small–medium companies can be developed using this order analysis technique. Further, we validate the effectiveness of these rules developed by order analysis, in comparison with those built by other methods. The rules developed by the proposed method outperform those of other traditional methods in effectively screening the bankrupted firms.