融合地理信息的兴趣点推荐

摘要

智能移动终端在大学校园内的迅速普及，使得校园无线局域网被广泛部署于图书馆、食堂、教学楼、宿舍等区域，推进了信息化校园的建设，为师生的校园生活提供了极大的便利，与此同时大学校园积累了大量基于位置的社交网络(LBSNs, Location-Based Social Networks)数据，如何利用LBSNs进行学生兴趣点(POI, Point-of-Interest)推荐已经成为一个研究热点。校园地理信息的高可用性为提高个性化兴趣点推荐的性能提供了机会，提高POI推荐性能，应解决两个主要挑战：首先，如何利用地理信息来获取用户个人信息、地理坐标和位置流行度等信息；然后如何将地理信息纳入推荐算法中。本文使用一种基于校园地理信息的Logistic矩阵分解(CGLMF, Campus Geographic Information based Logistic Matrix Factorization) POI推荐算法，该算法利用学生个人信息和校园地理信息，通过考虑学生的主要活动区域和该区域每个POI的相关性，提出一种有效的地理信息模型，然后将地理信息模型融合到Logistic矩阵分解中以此提高POI推荐性能，在校园真实学生Wi-Fi签到数据集上进行实验，结果表明该方法优于其他POI推荐方法。 With the rapid spread of smart mobile terminals, university campus wireless local area networks have been widely deployed in libraries, cafeterias, teaching buildings, dormitories and other areas, which has promoted the construction of information-based campuses and provided a great con-venience for teachers or students on campus. At the same time, the university campus has accu-mulated a large number of location-based social networks (LBSNs) data. How to use LBSNs for Point-of-Interest (POI) recommendation has become a research hotspot. The high availability of campus geographic information provides an opportunity to improve the performance of person-alized POI recommendations. However, there are two main challenges which should be addressed: First, use geographic information to obtain user personal information, geographic coordinates, and location popularity, etc.; second, incorporate the geographic information into the recommendation algorithm. This paper uses a Campus Geographic Information Based Logistic Matrix Factorization (CGLMF) POI recommendation algorithm. This algorithm uses student personal information and campus geographic information. An effective geographic information model is proposed by considering the student’s main activity area and the relevance of each POI in the area. Then the geographic information model is integrated into the logistic matrix decomposition to improve the performance of POI recommendation. Experimental results on the real-world students Wi-Fi check-in dataset on campus that the proposed approach outperforms other POI recommendation methods.