摘要
目前基于协同过滤的兴趣点推荐算法能够获得较好的推荐效果,但是当用户外出远离其常驻地时,推荐效果急剧下降,主要原因是用户的签到记录主要集中在其常驻地周围,而对其他兴趣点的签到行为较少,此时不能准确计算用户兴趣。因此提出了一种基于主题模型的兴趣点推荐算法,在推荐过程中同时考虑了用户的偏好分布和兴趣点的主题分布,使得当用户在新的兴趣点时,也能获得较好的推荐。实验证明,该方法不仅能够缓解推荐数据的稀疏性问题,而且与其他方法相比有更高的推荐准确率。
The existing point of interest recommendation algorithm based on the collaborative filtering can obtain good effect, but when users are far away from their residence, the recommend effect has fallen sharply. The main reason is that user's check-ins mainly concentrate around its residence, and has less activity history in other locations, which results in not accurately calculating the user interest at this time. This paper proposes a point of interest recommendation method based on topic model, which can give consideration to both personal interest and point of interest's preference in order to get satisfactory recommendation lists in a new city. Experiments show that the methods can not only ease data sparseness problem to a certain extent, but also outperform other methods.
出处
《计算机工程与应用》
CSCD
北大核心
2018年第4期154-159,共6页
Computer Engineering and Applications
基金
国家自然科学基金(No.61272277)
关键词
位置社交网络
兴趣点推荐
协同过滤
location-based social networks
location recommendation
collaborative filtering
