Extraction of design variables using collaborative filtering for interactive genetic algorithms

Abstract

Interactive Genetic Algorithm (iGA) is one of evolutionary computations in which the design candidates are evaluated by human. Using iGA, the sensibility and subjective feelings of humans can be optimized by learning the user's evaluation of presented individuals. In this research, iGA was applied to product recommendation on shopping sites. One of the most difficult points to be addressed in construction of a product recommendation system is to taking a long time to extract and assign values to design variables from all of the actual products on the site. It is also difficult to define product design variables appropriately. To address these problems, we propose a method to generate design variables automatically based on a lot of users' preference data on the Web. We constructed the design variables using the relevance of products obtained by Collaborative Filtering and discussed them. Through the simulation experiments, the effectiveness of the proposed method is discussed.