Abstract:

POI recommendation is an important personalized service in Location-Based Social Network（LBSN）, which can help users discover POIs and improve the quality of information services. Aiming at the problem of data sparsity in POI recommendation, this paper proposes a POI recommendation algorithm combining social relationship and local geographic factors. This algorithm measures user similarity based on the common check-in and distance relationships among users in the social relationship, and builds a social model through user collaborative filtering. A local activity area for each user is divided, and the sign-in correlation is analyzed to establish a local geographic factor model. Based on weighted matrix decomposition, users’own preferences are mined, and social relationships and local geographic factors for POI recommendation are integrated. Experiments on the Gowalla dataset show that the proposed POI recommendation algorithm has higher accuracy and recall rate than other methods, which can effectively alleviate the problem of data sparsity and improve recommendation performance.

Key words: Location-Based Social Networks（LBSN）, Point-Of-Interest（POI） recommendation, social relationship, local geographic factors, weighted matrix factorization

摘要：

兴趣点（Point-Of-Interest，POI）推荐是基于位置社交网络（Location-Based Social Network，LBSN）中一项重要的个性化服务，可以帮助用户发现其感兴趣的[POI]，提高信息服务质量。针对[POI]推荐中存在的数据稀疏性问题，提出一种融合社交关系和局部地理因素的[POI]推荐算法。根据社交关系中用户间的共同签到和距离关系度量用户相似性，并基于用户的协同过滤方法构建社交影响模型。为每个用户划分一个局部活动区域，通过对区域内[POIs]间的签到相关性分析，建立局部地理因素影响模型。基于加权矩阵分解挖掘用户自身偏好，并融合社交关系和局部地理因素进行[POI]推荐。实验表明，所提出的[POI]推荐算法相比其他方法具有更高的准确率和召回率，能够有效缓解数据稀疏性问题，提高推荐质量。

关键词: 位置社交网络, 兴趣点推荐, 社交关系, 局部地理因素, 加权矩阵分解