基于用户多兴趣和社交网络的个性化推荐研究

发布时间：2018-05-25 00:26

  本文选题：用户兴趣建模 + 社交网络　； 参考：《天津大学》2016年硕士论文

【摘要】：随着信息技术的发展和经济社会信息化进程的加快,电子商务进入繁荣发展时期。电子商务中的商品规模急剧增加,用户发现满意商品的困难增大,“信息过载”等问题日益严重。个性化推荐技术能够基于用户在网上的行为挖掘用户兴趣,从而主动地向用户推荐其可能感兴趣的商品。准确获取用户的兴趣,是个性化推荐的基础,在推荐系统中发挥着核心作用。用户兴趣一般不是单一的,而是呈现多样化状态。此外,随着社交网络的发展,有不少研究指出用户和其社交好友之间存在相似的兴趣,用户也更加信赖来自社交好友的推荐。如果能够结合社交网络信息,推荐性能可能会得到进一步的提升。本文结合用户多兴趣模型和社交网络相关理论,应用神经网络技术,针对用户兴趣建模展开研究。本文介绍了个性化推荐的原理与技术,综述了个性化推荐的研究现状;提出了一种用户兴趣模型,充分考虑用户的多种兴趣,将其与协同过滤算法进行结合后能够针对用户的不同兴趣分别进行推荐;在多兴趣模型的基础上,结合社交网络信息,引入社交好友兴趣来增强用户兴趣模型;在多个数据集上进行实验验证,实验结果表明本文算法推荐准确性较高、多样性较强,并且在一定程度上能够有效缓解用户冷启动问题。
[Abstract]:With the development of information technology and the acceleration of economic and social informatization process, electronic commerce has entered a prosperous period. In electronic commerce, the scale of goods increases rapidly, the difficulty of users finding satisfied goods increases, and the problem of "information overload" becomes more and more serious. Personalized recommendation technology can be based on the behavior of users on the Internet mining user interest, so as to actively recommend to the user may be interested in the goods. Accurate acquisition of user interest is the basis of personalized recommendation and plays a central role in the recommendation system. User interest is generally not a single, but presents a diversified state. In addition, with the development of social networks, many studies have pointed out that users and their social friends have similar interests, and users rely more on recommendations from social friends. If you can combine social network information, recommendation performance may be further improved. Based on the theory of user multi-interest model and social network, this paper applies neural network technology to research user interest modeling. This paper introduces the principle and technology of personalized recommendation, summarizes the research status of personalized recommendation, and proposes a model of user interest, which fully considers the various interests of users. Based on the multi-interest model and the social network information, the user interest model can be enhanced by introducing the social friend interest. Experimental results on multiple data sets show that the proposed algorithm has high accuracy and diversity, and can effectively alleviate the cold start problem of users to a certain extent.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.3
，

本文编号：1931261