Selection of Parameters Based on Fuzzy Extension Matrix
Jinghong Wang, Jiaomin Liu · 2006
Fuzzy extension matrix (FEM) inductive learning is an important method that generates knowledge from cases. Compared with conventional extension matrix techniques, it is more powerful and practical to handle with ambiguities in classification problems. Rule extraction from fuzzy extension matrix involves three parameters alpha, beta and gamma. These parameters play an importation role in the entire process of rule extraction based on FEM. They greatly affect the computation of fuzzy entropy and extract rules, however those important parameter value are usually estimated based on users by domain knowledge, personal experience and requirements. This paper introduces an approach to optimization of the three parameters based GA, and provides some theoretical support of directly selection of the parameter values through experiment. The main contributions of this paper are as follows: by combining GA and local search methods, we can get the reasonable parameters. Five data sets from the UCI machine learning database are employed in the study. Experimental results and discussions are given