Extracting Preference Rules Using Kansei Retrieval Agents with Fuzzy Inference

Abstract

This study examines a Kansei retrieval agent (KaRA) model based on fuzzy reasoning in terms of optimizing fuzzy rules. The KaRA model learns the user’s preferences based on the user’s sensory evaluation, and retrieves what the user wants from a large amount of data. The KaRA model based on fuzzy reasoning has information on membership functions and fuzzy rules. Previous studies have demonstrated the effectiveness of the KaRA model in terms of learning the user’s evaluation criteria by optimizing the membership function of fuzzy reasoning using numerical simulation. However, the fuzzy rules of the model have not been optimized. By optimizing fuzzy rules, the KaRA model can acquire the user’s sensibility evaluation information in linguistic expressions (fuzzy rules). Therefore, we confirmed the effectiveness of fuzzy rule optimization in the KaRA model. We conducted numerical simulations using pseudo-users and experiments with real users. Consequently, we examined the effectiveness of fuzzy rule optimization in the KaRA model.