earticle

영어

In recent years, several researches have been conducted on spectral clustering to classify non-linear data in various applications. Considering the effect of selecting the appropriate eigenvectors on spectral clustering performance; various methods have been proposed weighting and ranking features. However, these methods can independently evaluate the impact of each eigenvector. Nevertheless, it is possible that several eigenvectors have duplicate or inadequate information on some clusters. Thus, we have presented a new method for finding the optimal combination of eigenvectors by several different evaluation criteria. In order to detect simultaneously the optimum condition in various criteria, the multi-objective genetic algorithm is applied. Findings of performed experiments on datasets with various features demonstrate a resounding success in the proposed method.