基于潜在兴趣和地理因素的个性化兴趣点推荐研究

PC-Geo: Potential Check-ins and Geographical Model based Point-of-Interest Recommendation

兴趣点推荐是基于位置的社交网络(LBSN)中一个重要的个性化服务。由于用户数据的稀疏性,该研究面临着严重的冷启动问题。为了应对这一挑战,提出一个能够利用情景信息来缓解冷启动,提高推荐效率的框架。首先,使用情境信息(空间信息、时间信息、社交信息)来为目标用户定义相关用户;其次,利用相关用户的历史签到数据,使用该算法学习目标用户的潜在兴趣,缓解签到矩阵的稀疏性。接着,把地理信息和矩阵分解的结果融合,最后完成对用户的推荐。

Point-of-Interest（POI） recommendation is an important application in location-based social networks（LBSNs）. However, the sparseness of users＇ brings big challenges for the study of POI recommendation. To tackle this problem, proposes a framework named PC-Geo that utilizes contextual information to POI recommendation to relieve the cold-start problem and improve the efficiency of recommendation. First, uses three types of contextual information（spatial information, temporal information and social information） to define correlated users for a user. Second, designs an algorithm that combines user interest and geographic information to learn a set of potential POIs for users. Finally,combines the geographical information and the result of the matrix factorization to recommend unvisited locations for users.