基于位置相关数据的用户行为分析及预测

发布时间：2019-01-28 18:45

【摘要】：随着互联网的飞速发展,社交网络成为了覆盖群体最广、社会影响最大、商用价值最高的Web2.0服务。移动终端设备上定位技术的成熟与提升,极大地改善了基于位置的服务的质量。社交网络与基于位置的服务之间的融合诞生出了一种新型的社交网络服务——基于位置的社交网络。基于位置的社交网络的显著特点是能够为用户记录带有时间戳的位置信息。借助位置信息,基于位置的社交网络的出现使虚拟世界与现实世界之间建立了更加紧密的关系。目前,基于位置的社交网络已经成为社交研究领域的一个新方向,它为相关的研究提供了大量的包含了时间、空间和社交关系的用户位置签到数据。基于位置的社交网络的各类服务,也为用户行为预测和推荐系统等相关研究提供了更多的应用点。本文的研究对象是基于位置的社交网络中用户的位置签到数据,以及相关的用户注册位置信息、线上社交关系网络等数据内容。通过对基于位置社交网络中的用户的线上行为特征的分析,发现了用户位置签到行为在时间、空间与社交关系上的多种特征,并依此建立用户行为特征模型。在相应的应用场景中,本文提出了一套个性化位置推荐系统,,给出了系统的详细设计。该系统根据用户特征模型,预测用户潜在的兴趣位置,为用户推荐出行建议,提升了用户在基于位置的社交网络中享受位置推荐服务的体验。本文基于用户线上同行行为的定义,提出了面向同行群组的位置群推荐算法。在多人同行场景下,与面向个人的推荐算法相比,提升了推荐算法的准确度。该算法,融合了本文提出的用户个性化位置推荐算法。测试了不同聚合方案中位置群推荐算法的性能,并基于分布式计算平台Spark进行了系统实现,提高了面向同行群组的位置群推荐性能。基于理论分析模型,本文提出了一个基于用户位置行为特征的远程提问系统。最终,针对本文的理论与实验两个方面给出了改进方向。
[Abstract]:With the rapid development of the Internet, social network has become the most widely covered group, the largest social impact, the highest commercial value of Web2.0 services. The maturity and improvement of location technology in mobile terminal devices greatly improve the quality of location-based services. The fusion of social networks and location-based services has given birth to a new type of social-networking service, location-based social network. A prominent feature of location-based social networks is the ability to record timestamp location information for users. With the help of location information, the emergence of location-based social networks makes the virtual world and the real world more closely related. At present, location-based social network has become a new direction in the field of social research. It provides a large number of user location check-in data including time, space and social relations for related research. The various services of location-based social networks also provide more application points for user behavior prediction and recommendation systems. The object of this paper is the location check in data of the user in the location-based social network, and the related information of the user registration location, the online social relationship network and other data content. By analyzing the online behavior characteristics of users in location-based social networks, this paper finds out a variety of features of user location-check-in behavior in time, space and social relations, and establishes a user behavior feature model. In the corresponding application scenario, this paper presents a personalized location recommendation system, and gives the detailed design of the system. According to the user characteristic model, the system predicts the potential location of interest of users, recommends travel advice for users, and improves the experience of users enjoying location recommendation services in location-based social networks. Based on the definition of peer behavior on user line, this paper proposes a location group recommendation algorithm for peer groups. Compared with the personal-oriented recommendation algorithm, the accuracy of the recommendation algorithm is improved in the multi-person peer scenario. This algorithm combines the user personalized location recommendation algorithm proposed in this paper. The performance of location group recommendation algorithm in different aggregation schemes is tested and implemented based on distributed computing platform Spark to improve the performance of position group recommendation for peer groups. Based on the theoretical analysis model, this paper presents a remote questioning system based on user location behavior characteristics. Finally, the direction of improvement is given in view of the theory and experiment of this paper.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.3

【参考文献】