Development of multi-fuzzy classifiers using multi-objective genetic algorithm for medical diagnosis

Semra GÜNGÖRFIRAT UNIVERSITYGRADUATE SCHOOL OF NATURAL AND APPLIED SCIENCESDEPARTMENT OF COMPUTER ENGINEERINGMASTERS?Development Of Multi-Fuzzy Classifiers Using Multi-Objective Genetic Algorithm ForMedical Diagnosis?In this thesis, data mining based multi-fuzzy classifiers are developed by using multi-objectivegenetic algorithms for medical diagnosis. The superiority of multi-objective genetic algorithms overclassic approaches is that many non-dominated solutions, which represent classifiers in this thesis, canbe obtained by their single run. That is, multiple classifiers are obtained by applying an multi-objective genetic algorithm to training pattern just once.The main advantage at our method is that a large number of tradeoff fuzzy classificationsystems can be extracted with respect to conflicting objectives: accuracy maximization, ruleminimization and total length minimization.Experimental results conducted on two well-known medical data set having quantitativeattributes, hepatitis and diabetes, demonstrate that high classification performance on test pattern canbe obtained from the non-dominated fuzzy classifier.