Using Chi-square Matrix to Strengthen Multi-objective Evolutionary Algorithm

Many complex engineering problems have multi-objectives where each objective is conflicting with others. However, a lot research Jiradej Ponsawat et al. 2 works in optimization by Competent Genetic Algorithm are focused on single objective methods. These algorithms work very well for single objective problems but stumble when trying to discover a large number of solutions naturally occurred in multi-objective problems. There are many multi-objective problems where solutions share common characteristic, for example decomposable multi-objective problems. This characteristic can be exploited to identify and compose these common structures. This work proposes to apply the concept of Building Blocks to improve evolutionary algorithms to tackle such problems. Building Block Identification algorithm is used to guide the crossover operator in order to maintain good Building Blocks and mix them effectively. The proposed method is evaluated by using Building Block Identification guided crossover in a well-known Genetic Algorithm to solve multiple-objective problems. The result shows that the proposed method is effective. Moreover, it obtains a good spread of solutions even when the Building Blocks are loosely encoded.