A fuzzy genetics-based machine learning method for designing linguistic classification systems with high comprehensibility

In this paper, we examine the performance of two fuzzy genetics-based machine learning approaches to the design of linguistic classification systems. One is the Michigan approach in which each linguistic rule is coded as a string (i.e., an individual is a single linguistic rule). The other is the Pittsburgh approach in which a set of linguistic rules is coded as a string (i.e., an individual is a rule-based classification system). After demonstrating advantages and disadvantages of each approach, we combine these two approaches into a hybrid algorithm