An Improved Fuzzy Genetics-Based Machine Learning Algorithm for Pattern Classification

This paper presents an improved version of the hybrid fuzzy genetics-based machine learning algorithm [3] for pattern classification. We extend the original fuzzy rule form with a single consequent class to the form with multiple consequent classes for the reason of being more general in most cases. The original fuzzy reasoning with a single winner rule is also replaced with a weighted vote method accordingly. Besides, we cancel the step of rule optimization by the Michigan-style algorithm and add a heuristic procedure to speed up the algorithm. Experimental results show that our method produces better classification results and converges more quickly than the original version.