多特征融合的兴趣点推荐算法

基于位置社交网络的兴趣点推荐越来越受到工业界和学术界的关注。由于用户签到数据集的稀疏性以及签到地理位置的聚集性，使得目前的推荐算法效率普遍不高，特别是当用户外出到新的地点时，推荐效果更是急剧下降。因此本文提出了一种基于用户-区域-内容主题的多特征联合推荐算法（UCRTM），以隐主题模型为基础，在统一的框架下利用隐含因子关联性融合了用户的偏好、兴趣点的内容以及兴趣点所属地理区域主题等信息来进行推荐，使得用户无论身处何地，都能获得理想的推荐服务。本文在两种真实的数据集上进行了实验，结果表明该方法不仅能够克服数据的稀疏性以及弱语义性等问题，而且与其他方法相比具有更高的推荐准确率。

A point of interest recommendation algorithm based on multi-feature fusion

The point of interest recommendation service is receiving increasing attention from the industry and academia. The sparsity of users’ activity history datasets and aggregation of geological position prevent the current recommendation algorithm efficiency from being high, and especially, when a user goes out to a new city, the recommendation effect will fall sharply. Therefore, this paper presents a user-content-region topic model based on a joint recommendation algorithm, considering to the user’s preferences, the content of the point of interest, and the geographical area, making users obtain an ideal recommendation service irrespective of their location. An experiment was carried out on two real datasets, and the results show that this method can not only overcome problems such as data sparseness, weak semantic performance, but also has a higher recommendation accuracy compared with other methods.