基于用户兴趣的微博推荐方法研究

发布时间：2018-06-02 04:21

本文选题：用户兴趣 + 用户标签　；参考：《山东师范大学》2017年硕士论文

【摘要】：社交网络中的微博平台,近年来得到了广大用户的喜爱和关注。据了解,每天都会有不计其数的新用户加入该平台,并在平台上留下成千上万条信息。面对海量的微博信息,用户总是在不停地寻找与自己兴趣相一致的信息,那么如何从这些信息中发现用户的兴趣,并向其推荐感兴趣的微博,成为目前研究的一个热点问题。本文正是以此为出发点,针对微博推荐算法中所存在的问题进行相关的研究。首先,针对在用户兴趣挖掘阶段中存在的准确率不高的问题,本文提出了一种基于标签更新的微博用户兴趣挖掘算法;其次,针对微博推荐阶段中存在的冷启动问题,本文提出了一种融合标签与人工蜂群的微博推荐算法;最后,利用上述两种算法设计与实现了微博推荐原型系统。本文所研究的具体工作如下:(1)研究用户兴趣挖掘算法,提出一种基于标签更新的微博用户兴趣挖掘算法。首先,根据标签的多种特征,用户建立自身的初始兴趣;其次,利用用户关注人与用户间的相似度、用户关注人自身的影响力和用户关注人与用户间的亲密度三种关系计算标签的更新强度;最后,根据标签的更新规则对标签进行更新建立用户兴趣模型。该方法在准确率和召回率方面都有一定的提高,说明运用该方法表示用户兴趣具有一定的有效性。(2)研究微博推荐算法,提出一种融合标签和人工蜂群的微博推荐算法。首先,对用户标签信息进行定义;其次,利用已定义的标签权重、标签偏好和标签与微博中词语的相似度三种变量来构造人工蜂群中的适应度函数;最后,利用人工蜂群算法的搜索策略,搜索出具有最优适应度值的微博向用户进行推荐。该方法不仅可以解决推荐算法中的冷启动问题,而且对提高推荐算法的准确性也具有良好的效果。(3)设计与实现基于用户兴趣的微博推荐原型系统的。以上述两种算法为理论基础,分析和详细设计系统中所需的各个模块和流程,并最终实现基于用户兴趣的微博推荐原型系统,以供用户及时发现并找到自己所喜爱的微博信息。
[Abstract]:In recent years, the Weibo platform in the social network has been loved and concerned by the majority of users. Countless new users join the platform every day and leave thousands of messages on it. In the face of massive Weibo information, users are always looking for information consistent with their own interests, so how to find the interest of users from these information and recommend interested Weibo to them has become a hot issue. In this paper, we study the problems in Weibo recommendation algorithm. Firstly, aiming at the problem that the accuracy of user interest mining is not high, this paper proposes a new Weibo user interest mining algorithm based on tag updating. Secondly, aiming at the cold start problem in Weibo recommendation phase, In this paper, a Weibo recommendation algorithm combining tag and artificial bee colony is proposed, and finally, the prototype system of Weibo recommendation is designed and implemented by using the above two algorithms. The main work of this paper is as follows: 1) A new Weibo user interest mining algorithm based on tag update is proposed. First of all, according to the various features of the label, the user establishes his own initial interest; secondly, the user focuses on the similarity between the user and the user. The user pays attention to the influence of the person and the user pays close attention to the relationship between the user and the user to calculate the update intensity of the label. Finally, the user interest model is established according to the update rules of the label. This method has some improvement in accuracy and recall rate. It shows that it is effective to use this method to express user's interest. (2) to study the Weibo recommendation algorithm, and to propose a Weibo recommendation algorithm combining tag and artificial bee colony. Firstly, the user tag information is defined; secondly, the fitness function in artificial bee colony is constructed by using the defined tag weight, label preference and label similarity with words in Weibo. Using the search strategy of artificial bee colony algorithm, the Weibo with the optimal fitness value is searched for recommendation to the user. This method can not only solve the cold start problem in the recommendation algorithm, but also improve the accuracy of the recommendation algorithm. It has a good effect in designing and implementing the Weibo recommendation prototype system based on the user's interest. Based on the above two algorithms, this paper analyzes and designs each module and flow of the system in detail, and finally realizes the prototype system of Weibo recommendation based on user's interest, so that users can find and find their favorite Weibo information in time.
【学位授予单位】：山东师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.3

【参考文献】