Improving Effectiveness of Recommendation List and its Evaluation

This paper proposes to improve the effectiveness of recommendation items list by taking into account familiarity among customers, in the E-commerce domain. Recently, many online shops are attracted to push type information provision that recommends their items to a customer. Collaborative filtering is one of the representative techniques for such purpose. However, it has mainly discussed recommendation precision, and satisfaction for customers has been not always sufficient. In general, it is possible for a customer to be interested by themself in the item which a familiar person likes. In the present paper, we propose introduction of taking into account familiarity among customers into the conventional collaborative filtering, and try to improve the effectiveness of recommendation list. The experiment results using our pilot system have shown that our proposition would be more effective than the general collaborative filtering. Key-Words: E-Commerce, item recommendation, collaborative filtering, effectiveness, familiarity, television program