閱讀全文 | |
篇名 |
GRUIFI: A Group Recommendation Model Covering User Importance and Feature Interaction
|
---|---|
並列篇名 | GRUIFI: A Group Recommendation Model Covering User Importance and Feature Interaction |
作者 | Jingwei Zhang、Jing Cheng、Ya Zhou、Qing Yang |
英文摘要 | Group recommendation derives from a phenomenon that a group with similar interests have formed various communities, which creates the requirements that a group of users in one community want to share personalized services. Different from traditional recommendations that focus on individuals, group recommendation needs to consider the differences in preference of group members. How to build a proper model for group members to aggregate different preferences is still a challenging problem: (1) the influence of group members is quite different; (2) a user decision is directly or indirectly influenced by other members in the same group. This paper proposed a Group Recommendation model covering User Importance and automatic Feature Interaction (GRUIFI), which can model interaction data of group member and learn group potential preference representation. Our model exploits an attention mechanism to obtain the weights of group members that represent user importance, and those dynamic user weights are integrated to learn a group representation. Then we design a neural network that combines the multi-head attention to automatically learn fine-grained interactions between groups and items, and further capture the interdependency between group members. Finally, the experiments on the two real-world datasets show that GRUIFI performs significantly better than baseline methods. |
起訖頁 | 1141-1153 |
關鍵詞 | Group recommendation、Preference aggregation、High-order feature interactions、Group representation |
刊名 | 網際網路技術學刊 |
期數 | 202109 (22:5期) |
出版單位 | 台灣學術網路管理委員會 |
DOI |
|
QR Code | |
該期刊 上一篇
| KFPA Monocular Ranging Algorithm Design and Application in Mobile edge Computing |
該期刊 下一篇
| Linear and Lossy Identification Schemes Derive Tightly Secure Multisignatures |