Using Rough Sets Theory and Minimum Description Length Principle to Improve a β-TSK Fuzzy Revision Method for CBR Systems

摘要

This paper examines a fuzzy logic based method that automates the review stage of a 4-step Case Based Reasoning system and aids in the process of obtaining an accurate solution. The proposed method has been derived as an extension of the Sugeno Fuzzy model, and evaluates different solutions by reviewing their score in an unsupervised mode. In addition, this paper proposes an improvement of the original fuzzy revision method based on the reduction of the original set of attributes that define a case. This task is performed by a feature subset selection algorithm based on the Rough Sec theory and the minimum description length principle.