A genetic fuzzy approach for rule extraction for rule-based classification with application to medical diagnosis
Rahil Hosseini, Tim J. Ellis, Mahdi Mazinani, Jamshid Dehmeshki · 2011
Abstract — Rule extraction is an important task in knowledge discovery from imperfect training dataset in uncertain environments such as medical diagnosis. In a medical classification system for diagnosis, we cope with expensive or lack of expert knowledge in the design of the classifier. This paper presents an evolutionary fuzzy approach for tackling the problem of uncertainty in the process of rule extraction for classification. A type-2 fuzzy logic (T2FL) in combination with genetic algorithm is proposed for modeling uncertainty a long with rule extraction process. The approach has been applied on the Wisconsin breast cancer diagnostic dataset (WBCD) with an average classification accuracy of 96 % which is competitive with the best results to date. Furthermore, the proposed T2FLS maintains the tradeoffs between interpretability and accuracy by producing the most comprehensive rule set with only one linguistic term per rule in compare to other methods proposed for the Wisconsin breast cancer diagnosis. Keywords—Rule extraction, type-2 fuzzy logic system, Genetic algorithm, Classification,