Recommendation System Based on Unexpectedness Index that Balances Estimated Purchase Probabilities and Predicted Evaluation Values

DOI Web Site
  • SEKIGUCHI Ayumi
    Department of Industrial and Management Systems Engineering, School of Creative Science and Engineering, Waseda University
  • NINOHIRA Masato
    Department of Industrial and Management Systems Engineering, School of Creative Science and Engineering, Waseda University
  • MIKAWA Kenta
    Department of Information Science, Faculty of Engineering, Shonan Institute of Technology
  • GOTO Masayuki
    Department of Industrial and Management Systems Engineering, School of Creative Science and Engineering, Waseda University

Bibliographic Information

Other Title
  • 推定購買確率と予測評価値をバランスする意外性指標に基づく推薦システム
  • スイテイ コウバイ カクリツ ト ヨソク ヒョウカチ オ バランス スル イガイセイ シヒョウ ニ モトズク スイセン システム

Search this article

Description

<p>The development in information technology enables to store various kinds of data e.g.) purchase history data and evaluation history data for purchased items and so on. To make use of these data, the recommender system that recommends items that match customer preferences is widely used. Generally, the recommender system recommends items that a customer has not purchased and seems to purchase with high possibility. However, those items are likely purchase without recommendation, therefore, it needs to be recommended the item whose serendipity is high in order to improve customer satisfaction. From the above discussions, we propose the unexpected index that balances estimated purchase probabilities and predicted evaluation values, and recommender system using those indices. When constructing the recommender system, we use the Aspect Model, which is widely known as one of probabilistic latent class model. To verify the effectiveness of our proposed method, we conduct simulation experiments using benchmark data set of recommender system.</p>

Journal

Related Projects

See more

Keywords

Details 詳細情報について

Report a problem

Back to top