基于位置社交网络的上下文感知的兴趣点推荐

发布时间：2018-04-13 14:26

本文选题：基于位置的社交网络 + 兴趣点推荐　；参考：《计算机学报》2017年04期

【摘要】：随着基于位置社交网络(Location-Based Social Networks,LBSN)的快速发展,兴趣点(Point-of-Interest,POI)推荐为基于位置的服务提供了前所未有的机会.兴趣点推荐是一种基于上下文信息的位置感知的个性化推荐.然而用户-兴趣点矩阵的极端稀疏给兴趣点推荐的研究带来严峻挑战.为处理数据稀疏问题,文中利用兴趣点的地理、文本、社会、分类与流行度信息,并将这些因素进行有效地融合,提出一种上下文感知的概率矩阵分解兴趣点推荐算法,称为TGSC-PMF.首先利用潜在狄利克雷分配(Latent Dirichlet Allocation,LDA)模型挖掘兴趣点相关的文本信息学习用户的兴趣话题生成兴趣相关分数;其次提出一种自适应带宽核评估方法构建地理相关性生成地理相关分数;然后通过用户社会关系的幂律分布构建社会相关性生成社会相关分数;另外结合用户的分类偏好与兴趣点的流行度构建分类相关性生成分类相关分数,最后利用概率矩阵分解模型(Probabilistic Matrix Factorization,PMF),将兴趣、地理、社会、分类的相关分数进行有效地融合,从而生成推荐列表推荐给用户感兴趣的兴趣点.该文在一个真实LBSN签到数据集上进行实验,结果表明该算法相比其他先进的兴趣点推荐算法具有更好的推荐效果.
[Abstract]:With the rapid development of location-based Social networks, Point-of-interest offers unprecedented opportunities for location-based services.Point of interest recommendation is a location-aware personalized recommendation based on contextual information.However, the extremely sparse user-point matrix poses a severe challenge to the research of point of interest recommendation.In order to deal with the problem of data sparsity, this paper uses the information of geography, text, society, classification and popularity of points of interest, and combines these factors effectively, and proposes a context-aware probability matrix decomposition algorithm called TGSC-PMF.Firstly, the potential Delikley assignment Latent Dirichlet allocation (LDAA) model is used to mine the text information related to the point of interest to learn the user's topic of interest and generate the interest correlation score.Secondly, an adaptive bandwidth kernel evaluation method is proposed to generate geographical correlation scores, and then social correlation scores are generated by power law distribution of user social relations.In addition, according to the classification preference of users and the popularity of interest points, the classification correlation scores are constructed. Finally, the probability matrix factorization model is used to integrate the relevant scores of interest, geography, society and classification effectively, using probabilistic Matrix factorization model (Probabilistic Matrix factorization / PMFN).Thus a recommendation list is generated to recommend points of interest to the user.The experimental results on a real LBSN check-in dataset show that the proposed algorithm is more effective than other advanced recommendation algorithms for points of interest.
【作者单位】：北京邮电大学计算机学院信息网络工程研究中心教育部重点实验室;
【基金】：国家科技重点支撑项目(2014BAK15B01)资助~~
【分类号】：TP391.3

【相似文献】