抄録
Interactive genetic algorithms(iGAs) are methods that acquire and analyse user preference based on user's subjective evaluation. IGA has been applied to various unimodal problems, such as parameter setting of a hearing aid and fashion design. On the other hand, the goal of this study is achieving iGA which also corresponds to multimodal preferences with equivalent fitness at the peaks. For example, when users select products on shopping sites, they have several types of preference trends at the same time. In this case, reflecting all the trends in product presentation leads to increased sales and consumer satisfaction. The dependency among design variables should be also considered toward each trend of preference. In this study, a new offspring generation method that enables efficient search even if user preference is multimodal and there are dependencies among design variables is proposed. The proposed method is capable to detect a multimodal preference using clustering, and spawn off-spring based on probabilistic model eliminating the dependencies among design variables by principal component analysis. The results of experiments indicate that the proposed method can generate offspring corresponding to multimodal preferences with equivalent fitness at the peaks and dependencies among design variables. However, there are few subjects who have dependencies within their preferences.
