ADAPTED GEOMETRIC SEMANTIC GENETIC PROGRAMMING FOR DIABETES AND BREAST CANCER CLASSIFICATION

In this paper, we explore new Adapted Geometric Semantic (AGS) operators in the case where Genetic programming (GP) is used as a feature generator for signal classification. Also to control the computational complexity, a devolution scheme is introduced to reduce the solution complexity without any significant impact on their fitness. Fisher's criterion is employed as fitness function in GP. The proposed method is tested using diabetes and breast cancer datasets. According to the experimental results, GP with AGS operators and devolution mechanism provides better classification performance while requiring less training time as compared to standard GP.