Design of fuzzy hyperbox classifiers based on a two-stage genetic algorithm and simultaneous strategy

The fuzzy min-max (FMM) neural network can be regarded as a typical fuzzy hyperbox classifier that is designed in a sequential way, which leads to an input order drawback and overlap elimination limitation. In this paper, we propose a two-stage-based genetic algorithm (TGA) to construct blue a fuzzy hyperbox classifier (FHC) in a simultaneous way. The simultaneous method is realized by estimating all parameters of hyperboxes at one time rather than by separately determining the parameters of hyperboxes in a sequential way. In this paper, we propose a two-stage-based genetic algorithm to construct the fuzzy hyperbox classifier. The overall TGA consists of two stages, namely, the construction stage and optimization stage. The construction stage is aimed at designing the FHC structure, while the goal of the optimization stage is to further optimize the FHC structure. Using a two-stage genetic algorithm to directly construct a fuzzy hyperbox classifier can overcome the problem of input order and hyperbox overlap. The experimental results show that the proposed FHC yields higher classification accuracy in comparison with the stage-of-the-art FMMs reported in the literature.