社交网络好友推荐系统的设计与实现

发布时间：2018-10-30 17:57

【摘要】：在互联网迅猛发展的今天,人们获取信息的渠道越来越倾向于网络,结识朋友的途径也拓展到了社交网络上。社交的本质是人与人的交际往来,社交网络中每个用户都有各自的交际圈,并通过这个交际圈实现信息的传递、共享、交流。然而随着互联网络的发展和社交网络的演进,社交网络中用户群逐渐庞大,用户之间关系日益复杂,用户产生的数据量越来越多,种种因素无一不增加了用户寻找“志趣相投”的好友、建立自己交际圈的难度。在这种背景下,好友推荐系统应运而生,为目标用户推荐与其有着相似兴趣爱好的“朋友”。本文以微博这一社交网络的典型作为研究对象,对社交网络好友推荐系统进行设计与实现:首先,对微博用户数据进行采集和预处理,获得对于系统设计和实现有用的数据;其次,采用LDA主题模型方法对用户的微博内容进行分析,得到用户微博主题分布信息,并依据这些主题分布信息计算和表示出用户的兴趣偏好;再次,根据余弦相似度测度方法对不同用户兴趣之间的相似度进行计算,并选择与目标用户相似度最大的N个用户作为好友推荐结果呈现给目标用户;最后,采用准确率指标对该好友推荐系统进行评估,验证了系统对于推荐精度的提高。相对于传统的基于用户个性化标签、教育背景或地理位置等信息的好友推荐系统而言,本文提出的好友推荐系统,通过分析微博用户的历史微博数据,挖掘用户兴趣,并基于此进行好友推荐,因此对于用户兴趣的描述更具代表性,呈现给用户的推荐结果也更符合“志趣相投”这一标准。
[Abstract]:With the rapid development of the Internet, people's access to information is more and more inclined to the Internet, and the way to make friends is also extended to the social network. The essence of social communication is the communication between people. Each user in social network has his own social circle, and through this social circle, information transmission, sharing and communication are realized. However, with the development of the Internet and the evolution of the social network, the user group in the social network is gradually huge, the relationship between users is becoming more and more complex, and the amount of data generated by the users is more and more. All these factors make it more difficult for users to find friends with similar interests and establish their own social circle. In this context, the friend recommendation system emerges as the times require, and recommends "friends" with similar interests to the target users. Taking Weibo, a typical social network, as the research object, this paper designs and implements the friend recommendation system of social network. Firstly, we collect and preprocess the user data of Weibo, and obtain the useful data for the system design and implementation. Secondly, the LDA thematic model is used to analyze the Weibo content of the user, and then the user theme distribution information is obtained, and the interest preference of the user is calculated and expressed according to these thematic distribution information. Thirdly, according to the cosine similarity measure method, the similarity between different users' interests is calculated, and N users with the largest similarity to the target user are selected as the friend recommendation results to present to the target user. Finally, the accuracy index is used to evaluate the friend recommendation system, which verifies the improvement of recommendation accuracy. Compared with the traditional friend recommendation system based on user's personalized label, educational background or geographical location, the friend recommendation system proposed in this paper, through analyzing Weibo user's historical Weibo data, excavates user's interest. Therefore, the description of user's interest is more representative, and the recommendation result presented to the user is more in line with the standard of "like-minded".
【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】