基于语义位置和区域划分的兴趣点推荐模型

针对现有的位置社交网络研究工作对兴趣点相关的用户语义位置信息挖掘不够充分，且大多推荐算法忽略了兴趣点所在区域对推荐结果的影响，提出了一种新型兴趣点推荐模型（USTTGD）。首先采用分割时间的潜在狄利克雷分配主题模型（latent Dirichlet allocation，LDA），基于签到记录中的语义位置信息挖掘时间主题下的用户时间偏好，然后将兴趣点所处区域划分为网格，以评估区域影响；接着应用边缘加权的个性化PageRank（edge-weighted personalized PageRank，EwPPR）来建模兴趣点之间的连续过渡；最后将用户时间偏好、区域偏好和连续过渡偏好融合为一个统一的推荐框架。通过在真实数据集上实验验证，与其他传统推荐模型相比，USTTGD模型在准确率和召回率上有了显著的提升。

Point-of-interest recommendation model based on semantic location and regional division

According to the existing research work of location-based social network was not sufficient to mine the user semantic location information related to point-of-interest, moreover, most recommendation algorithms ignored the influence of the region of point-of-interest on the result of recommendation, this paper proposed a new recommendation model of point-of-interest called USTTGD. Firstly, it adopted the latent Dirichlet allocation(LDA) topic model of time division, and mined the user time preference under the time theme based on the semantic location information in check-in records. Then it devided the region of point-of-interest into grids to evaluate the regional influence. Next, it applied edge-weighted personalized PageRank(EwPPR) to model the successive transitions among point-of-interests. Finally, USTTGD model fused user time preference, regional preference and successive transition preference into a unified recommendation framework. Experimental results on real-world datasets show that USTTGD model achieves significantly enhance compared with other classical recommendation models on precision and recalling rates.