Spatial-temporal distance metric embedding for time-specific POI recommendation
With the growing popularity of location-based social networks (LBSNs), time-specific POI recommendation has become important in recent years, which provides more accurate recommendation services for users in specific spatio–temporal contexts. In this paper, we propose a spatio–temporal distance metr...
Saved in:
| Main Authors: | , , |
|---|---|
| 其他作者: | |
| 格式: | Article |
| 語言: | English |
| 出版: |
2018
|
| 主題: | |
| 在線閱讀: | https://hdl.handle.net/10356/89331 http://hdl.handle.net/10220/47032 |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
| 機構: | Nanyang Technological University |
| 語言: | English |
| 總結: | With the growing popularity of location-based social networks (LBSNs), time-specific POI recommendation has become important in recent years, which provides more accurate recommendation services for users in specific spatio–temporal contexts. In this paper, we propose a spatio–temporal distance metric embedding model (ST-DME) for time–specific recommendation, which exploits both temporal and geo-sequential property of a check-in to effectively model users’ time-specific preferences. Specifically, we divide timestamps of user’ check-ins into different time slots and adopt Euclidean distance rather than inner product of latent vectors to measure users’ preferences for POIs in a given time slot. We also apply a transition coefficient based on users’ most recent check-ins to incorporate geo-sequential influence in users’ check-in behaviors. A weighted pairwise loss with a hard sampling strategy is adopted to optimize latent vectors of users, POIs, and time slots in a metric space. Extensive experiments are conducted to demonstrate the effectiveness of our proposed method and results show that ST-DME outperforms state-of-the-art algorithms for time-specific POI recommendation on two public LBSNs data sets. |
|---|