Abstract Point-of-interests (POIs) recommendations aim at recommending locations to users on social platforms by analyzing their histories or combining other information. At present, the different granularity of factors (i.e. time, geography and sociability) are not thoroughly studied in existing works. To deal with this problem, we propose a two-stage coarse-to-fine POI recommendation algorithm based on tensor factorization and weighted distance kernel density estimation (KDE). At first stage, we take account of not only long-term preferences with sequential context, but also the crowd’s preferences to estimate the coarse user-category interest. And then a specific-designed weighted KDE with consideration of spatial distance is employed to determine the fine-grained user-location interest. To evaluate the proposed method, experiments are conducted on two real benchmark location-based social network (LBSN) datasets. And the results show that the proposed method outperforms the state-of-the-art methods and produces better POI recommendation.

中文翻译：

---

一种用于兴趣点推荐的粗到细用户偏好预测方法

摘要 兴趣点（POI）推荐旨在通过分析用户的历史或结合其他信息向社交平台上的用户推荐位置。目前，已有的作品对不同粒度的因素（即时间、地理和社交性）没有深入研究。为了解决这个问题，我们提出了一种基于张量分解和加权距离核密度估计（KDE）的两阶段粗到细POI推荐算法。在第一阶段，我们不仅考虑了具有连续上下文的长期偏好，而且还考虑了人群的偏好来估计粗略的用户类别兴趣。然后使用考虑空间距离的特定设计的加权 KDE 来确定细粒度的用户位置兴趣。为了评估所提出的方法，实验是在两个真实的基于基准位置的社交网络 (LBSN) 数据集上进行的。结果表明，所提出的方法优于最先进的方法并产生更好的 POI 推荐。