CYRCE: A pattern of knowledge representation and knowledge acquisition for decision on personal loans

摘要

In this paper, the authors propose a pattern of knowledge base structure for expert systems that address the now classical Banking application of decision making on personal loans. One of the difficulties in this area is to obtain a proper representation of the intuitive judgments and personal conjectures that credit analysts apply routinely in their work, since standard algorithmic modeling procedures can't be applied in that context. The proposed pattern of knowledge representation is based on a mixture of frames and rules which has proven to be effective in an application to the analysis of credit card requests in a large Mexican Commercial Bank. First, the conceptual framework of the system is introduced. Then, the knowledge base structure is described and some examples of representation and the application to a particular case are presented. Finally, some considerations are made about the knowledge acquisition process that is also a combination of (a) statistical techniques applied to on-field periodical surveys and (b) traditional interview methods. The possibility of a subsequent validation is also discussed.