Consistency in fuzzy rule based systems: an application in elementary school performance evaluation

Abstract

Performance evaluation and reform of public organizations is receiving increased attention. In many cases identification of the lowest performing organizations is an important part of performance based reforms. Programs to identify low performance would typically benefit from incorporation of expert judgement, provided that judgement could be applied in a verifiable and fair manner. Fuzzy rule based systems provide a mechanism for modeling expert judgement. However, translating that expertise into a consistent set of rules can be even more challenging for performance evaluation than it has been in engineering applications. We describe an approach that first creates a comprehensive and complete set of rules and then modifies those rules to prevent inconsistent results.