计算机工程 ›› 2018, Vol. 44 ›› Issue (7): 177-182.doi: 10.19678/j.issn.1000-3428.0047996

• 人工智能及识别技术 • 上一篇    下一篇

基于用户时空相似性的位置推荐算法

蒋翠清,疏得友,段锐   

  1. 合肥工业大学 管理学院,合肥 230009
  • 收稿日期:2017-07-18 出版日期:2018-07-15 发布日期:2018-07-15
  • 作者简介:蒋翠清(1965—),男,教授、博士生导师,主研方向为数据挖掘、大数据分析;疏得友,硕士研究生;段锐,博士研究生。
  • 基金项目:

    国家自然科学基金(71571059)。

Location Recommendation Algorithm Based on Spatial-temporal Similarity of User

JIANG Cuiqing,SHU Deyou,DUAN Rui   

  1. School of Management,Hefei University of Technology,Hefei 230009,China
  • Received:2017-07-18 Online:2018-07-15 Published:2018-07-15

摘要:

在位置推荐中用户签到数据具有稀疏性的特点,基于用户的协同过滤算法难以准确搜索邻近用户,从而影响推荐效果。针对该问题,分别将用户签到的时间信息与空间信息融入用户相似度计算中,提出考虑用户时空相似性的位置推荐算法。根据时间对用户签到行为的周期性影响,通过对用户签到矩阵按时间进行分割引入时间属性,设计一种时间相似性计算方法,并根据时间相似性对用户-地点-时间矩阵进行填补,缓解因时间分割导致的用户-地点-时间矩阵高稀疏问题。基于用户签到行为的空间聚集性,通过多中心聚类算法发现用户签到的活跃区域,结合用户对活跃区域的偏好以及未签到地点与活跃区域中心的距离,计算用户的空间相似性。在Foursquare数据集上的实验结果表明,与传统基于用户的协同过滤算法相比,该方法在准确率、召回率与F1-Measure评估值方面性能均有所提高。

关键词: 基于位置的社交网络, 位置推荐, 协同过滤, 时空相似性, 用户偏好

Abstract:

In the location recommendation,the result of finding users’ neighborhood is not good by using user-based collaborative filtering because of the sparseness of the check-in data,which results in the poor recommendation.Aiming at this problem,this paper proposes a location recommendation algorithm which considers user’s time and geographical similarity.It integrates the time and geography information of check-in data to compute user’s similarity.Based on the periodic influence of time on the user check-in behavior,the time attribute is introduced by dividing the user-location matrix over time.A time-similarity calculation method is designed,and the user-location-time matrix is filled according to the time similarity,which alleviates the problem of user-location-time sparseness caused by time segmentation.Based on spatial aggregation of user’s check-in behavior,the user spatial similarity is computed by combining the distance between the location and the center of user active region found by multi-center clustering algorithm and users’ preference of active region.Experiments is implemented on Foursquare dataset.Experimental result shows that the proposed algorithm achieves superior precision,recall and F1-measure compared with the traditional user-based location recommendation algorithm.

Key words: location-based social network, location recommendation, collaborative filtering, spatial-temporal similarity, user preference

中图分类号: