Interactive Recommender System to Estimate Personal User’s Kansei Model

The purpose of this research is to develop recommendation system reflecting individual user’s Kansei model. In the target contents have some keywords. The proposed method has following features. A recommendation problem is formulated as an optimization problem. The design space is defined with keywords of contents. The distance of each keyword is calculated by the network information which is developed by the whole contents in the system. The evaluation values of contents is derived by the interaction operation between the system and a user. The landscape of evaluation values in design space is called “Kansei Model” in this study. Using Kansei model information, an optimum point which is a recommendation content is extracted by interactive evolutionary computation (iEC). To analyze generated networks and investigate tendency of recommendation results by the proposed method, subjective experiment was performed by using large-sized product dataset. In the experiment result, it was confirmed that the proposed method could recommend products fitting subjective Kansei model. Keywords; recommender system, Kansei model, interactive evolutionary computation, keyword extraction