Study On The Generalization Ability Of Weighted Fuzzy Rules

In order to improve the capability of knowledge representation of FPRs, several parameters such as local weight, global weight have been incorporated into the FPRs, then we get the WFPR. Regarding the weights as adjustable parameters, and the capability of FPR for knowledge representation and reasoning can be improved, but it is a very difficult problem that how to get the values of these weights. The most existing approaches to weight learning are based on the further reduction of training error, which is really helpful to improve the training accuracy, but it is very likely to lead to an over-fitting and therefore seriously downgrades the GC of the FPR-based expert system. Firstly, this paper explores the relationship between the values of these weights and the GC. Secondly, this paper proposes a new weights learning scheme based on the well known fuzzy entropy maximization. Experimental results on a number of selected databases demonstrate the expected improvement of GA of the FPR-based expert systems.