Using Fuzzy Integral to Model Case-Base Competence

Modeling case-base competence is a crucial issue in the field of CBR. In fact it has led to a number of important developments in CBR recently, the most well known being in case deletion and case-base visualization and authoring support. This is highlighted by Smyth and Keane’s competence model (called the S-K model in this paper), which was shown to be efficient. In this paper, we argue that the model is not always a good predictor of competence, especially when problem distributions are nonuniform. Consequently, a new competence model based on fuzzy (non-linear) integral is proposed to address such a problem. The main idea is to repartition competence groups to ensure that the distribution of each group is nearly uniform, and more importantly to use the fuzzy integral with respect to a fuzzy measure (non-additive set function) to compute the overall competence. The interaction among new competence groups is considered to be reflected in the non-additive set function. The advantage of the newly proposed model is its high accuracy for predicting competence especially in the situation of non-uniform distributed case-base.