面向微博用户的内容与好友推荐算法研究与实现

发布时间：2018-10-17 22:11

【摘要】：随着网络信息的增加,信息过载现象日益严重。针对微博平台中的信息过载和推荐准确率低的问题,本文利用社交网络、奇异值分解、文本分类等技术对微博推荐效果进行改进。提出并介绍了三种推荐算法,并对三种推荐算法进行了深入的研究。本文的主要研究内容如下： (1)研究了基于社交网络的推荐算法,并在此基础上提出了基于社交网络和信任度的微博好友推荐算法。该算法针对社交网络推荐准确率低的问题,提出通过结合用户关注关系和用户行为来预测用户偏好的方法,并在此基础上为用户做好友推荐。 (2)针对微博中没有评分数据的问题,本文通过将用户行为转化为评分的方法进行用户偏好预测,并提出了基于奇异值分解的微博好友推荐算法,该算法通过对生成的用户评分矩阵进行奇异值分解来降低矩阵维数,并结合用户相似性和文本相似性进行偏好预测和推荐。 (3)由于微博中含有大量的文本,并且用户喜好的种类不唯一,因此本文提出了基于文本分类的微博内容推荐算法。该算法针对微博平台中存在的大量文本信息进行文本挖掘,通过文本分类对用户的偏好进行整理,分类的为用户进行内容推荐,从而提高推荐的准确率和召回率。最后通过实验测试表明三种推荐算法在准确率和召回率方面有了一定程度的提高。
[Abstract]:With the increase of network information, the phenomenon of information overload is becoming more and more serious. Aiming at the problem of information overload and low recommendation accuracy in Weibo platform, this paper uses social network, singular value decomposition, text classification and other techniques to improve the recommended effect of Weibo. Three recommendation algorithms are proposed and introduced, and three recommendation algorithms are studied in depth. The main contents of this paper are as follows: (1) the recommendation algorithm based on social network is studied, and on this basis, Weibo friend recommendation algorithm based on social network and trust degree is proposed. Aiming at the problem of low recommendation accuracy in social networks, this algorithm proposes a method to predict user preferences by combining user concern and user behavior. And on this basis to make friends recommendation for users. (2) in order to solve the problem that Weibo has no scoring data, this paper uses the method of converting user behavior to scoring to predict user preference. A friend recommendation algorithm of Weibo based on singular value decomposition (SVD) is proposed. The algorithm reduces the dimension of the generated user score matrix by singular value decomposition. And combined with user similarity and text similarity to predict and recommend preferences. (3) because Weibo contains a large number of texts, and user preferences are not unique, Therefore, this paper proposes Weibo content recommendation algorithm based on text classification. Based on the text mining of a large amount of text information in Weibo platform, the algorithm arranges the user's preference by text classification, and recommends the content for the user by classifying, so as to improve the accuracy and recall rate of recommendation. Finally, the experimental results show that the accuracy and recall rate of the three recommended algorithms have been improved to a certain extent.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.092;TP391.3

【参考文献】