| 张翔宇*******,张强***,吕明琪***.基于GPS轨迹挖掘的兴趣地点个性化推荐方法[J].高技术通讯(中文),2021,31(1):75~83 |
| 基于GPS轨迹挖掘的兴趣地点个性化推荐方法 |
|
| |
| DOI：10.3772/j.issn.1002-0470.2021.01.008 |
| 中文关键词: 个性化兴趣地点（POI）推荐； 基于位置社会网络； 全球定位系统（GPS）轨迹； 时空数据挖掘 |
| 英文关键词: personalized place of interest (POI) recommendation, location-based social network, global positioning system (GPS) trajectory, spatiotemporal data mining |
| 基金项目: |
| 作者 | 单位 | | 张翔宇******* | | | 张强*** | | | 吕明琪*** | |
|
| 摘要点击次数: 2145 |
| 全文下载次数: 1584 |
| 中文摘要: |
| 随着全球定位系统（GPS）等定位设备的普及，用户可方便地记录其GPS轨迹，这使得自动从用户的GPS轨迹中发现兴趣地点(POI)（如餐厅、商场、景点）并在用户之间进行推荐成为可能。因此，本文提出了一种基于GPS轨迹挖掘的兴趣地点个性化推荐方法。该方法与现有主流的兴趣地点推荐平台具有以下不同：首先，现有平台假设兴趣地点是事先已知的，而该方法通过一个层次化聚类算法从用户GPS轨迹中自动挖掘兴趣地点。其次，现有平台的推荐模式为平台向用户推荐，因此仅考虑了用户的偏好，忽略了用户之间的社交关联对推荐效果的影响。针对此问题，该方法基于用户交叠访问行为计算用户之间的社交信任度，基于用户访问行为的相似性计算其对兴趣地点偏好的相似度，在此基础上提出了一种能够融合用户信任度和相似度的评分算法。文本基于真实GPS轨迹数据对提出的方法进行了评测，实验结果表明，本文所提方法的综合推荐性能明显优于简单的基于访问数量的推荐方法、仅基于用户信任度的推荐方法及仅基于偏好相似度的推荐方法。 |
| 英文摘要: |
| With the increasing availability of global positioning system (GPS) enabled devices, users can easily record their GPS trajectories. It opens the possibility of discovering place of interests (POIs) (e.g., restaurants, malls, scenic spots) from GPS trajectories and recommending between users. Therefore, a method for personalized POI recommendation based on GPS trajectory mining is proposed. The proposed method differs from existing POI recommendation platforms in the following aspects. First, the existing platforms assume that the POIs are known in advance, while the proposed method automatically extracts POIs from GPS trajectories based on a hierarchical clustering algorithm. Second, the existing platforms only consider the users’ preference, ignoring the effect of social relations between users on the recommendation quality. Aiming at this problem, the proposed method estimates the social trust degree based on overlapping visiting activities and estimates the preference similarity on POIs based on the similarity of visiting activities. Then, it proposes an algorithm to fuse the social trust degree and POI preference similarity. The proposed method is evaluated based on real GPS trajectory dataset. The experiment results show that the overall performance of the proposed method outperforms the simple recommender based on visit counts, the recommender based only on social trust degree, and the recommender based purely on POI preference similarity. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
