Consistency Checking of Risk Policy Rules Using Meta-Rules: Axiomatic Approach to Optimal Credit Policy Evaluation and Trade Offs for Credit Policy Expert Systems

Abstract

This paper discusses a new approach to validating risk-related expert systems. The paper builds on existing knowledgebase verification and software testing literature to propose that risk related expert systems check for pareto optimal trade-offs in rule conditions along with the checks for ensuring contiguous numeric risk state space is covered by rules to ensure to gap in logic occurs. The construct of pareto optimal rules is novel and shown to be an effective and tractable approach used in production systems. In addition the pareto optimal property is form of common sense which financial credit related expert systems, which abound in industry, can benefit from as evidenced by the recent credit crisis. Ensuring that all rules are contiguous and pareto optimal within rule contexts adds to the expert system verification literature and is an approach that can increase early detection of defects along with ensuring sound credit decisioning for industrial expert systems. Thus the work has theoretical and practical value.