融合地理社交和时间序列信息嵌入排名位置推荐模型

在基于社会化媒体的位置推荐中，建模用户签到的位置序列建模十分必要。已有的相关算法大多都忽略了这样一个事实，即不同日子的签到序列表现出了不同的时间特征。为解决上述问题，提出一个地理社交时间序列嵌入排名（GSTSER）模型用于基于社会化媒体的位置推荐。该统一模型中的时间位置嵌入模型用于捕获序列中的上下签到信息以及不同日子的各种时间特征。同时，也提出了一种新的方法，根据地理—社交信息区分未访问的位置，将地理—社交影响纳入成对偏好排序方法。最后，基于一个统一的框架来结合这两种模型用于推荐位置。为了验证提出方法的有效性，在两个真实的数据集实验结果表明，GSTSER模型优于主流先进位置推荐算法。

Geo-social-temporal sequential embedding rank for point-of-interest recommendation

In social media-based location recommendation, it was necessary to model the location sequences of user''s check-in. The existing recommendation algorithms always ignored the fact that the check-in sequences on different days showed different time characteristics. In order to solve the above problem, inspired by the sequence context successfully modeled by the word2vec framework, this paper proposed a geo-social-temporal-sequential embedding ranking(GSTSER) model for Point-of-Interest recommendation. In GSTSER model, the time-location embedding model was used to capture the check-in information in the sequence as well as various time characteristics of different days. In the meantime, this paper also presented a new method of distinguishing unvisited locations based on geo-social information, which incorporates the effects of geo-social into a pairwise preference ranking model. Finally, the two models were combined to recommend locations based on a unified framework. Evaluation on two publicly datasets shows that our model performs significantly better than the state-of-the-art algorithms for social media-based location recommendation task.