Cold-start Recommendation Based on Deep Learning and Scenario-based Demographic Information

カテゴリ: 部門大会

論文No: SS7-1

グループ名: 【C】平成27年電気学会電子・情報・システム部門大会講演論文集

発行日: 2015/08/27

タイトル(英語): Cold-start Recommendation Based on Deep Learning and Scenario-based Demographic Information

著者名: Angze Li(Waseda University),Osamu Yoshie(Waseda University)

著者名(英語): Angze Li(Waseda University),Osamu Yoshie(Waseda University)

キーワード: コールドスタート|深層学習|制約付きボルツマンマシン|シナリオベース|Cold-start|Deep Learning|Restricted Boltzmann Machine|Scenario-based

要約(日本語): Collaborative filtering(CF) is the most popular approach in present recommender systems. But it suffers from some problems, one of which is well-known as cold-start problem due to a new user cannot be good recommended until he has a large number of existing records. For coping with this, our key idea is to combine traditional collaborative information with content feature. In this paper, our target is to improve the recommendation accuracy in new user cold-start situation. We describe a novel approach based on the traditional CF approach, the scenario-based demographic information and the Deep Learning method(Restricted Boltzmann Machines). Experimental results on movie-lens dataset shows that our new approach improves recommendation accuracy in new user cold-start recommendation tasks.

PDFファイルサイズ: 454 Kバイト