大数据下的基于主题模型的社交网络链接预测

Social Networking Link Prediction Based on Topic Model under Big Data

计算机技术和网络的发展使得数据呈爆炸式的涌现,社交媒体不断融入到人们的生活中,社会网络分析已成为研究的热点。随着大数据时代的到来,对社交网络链接算法研究产生巨大影响,原有的基于网络结构的预测方法已经渐渐不适应现状。因此,提出了一种基于主题模型的社交网络链接预测方法。首先以微博社交网络为数据源,将实验网络分为测试集和训练集;其次利用主题模型得到用户的主题特征,结合命名实体集和用户联系特征集合得到用户的兴趣特征相似性度量,加上网络结构相似性从而得到用户节点相似度,进而对社交网络链接进行预测;最终使用链接预测最常用的评价体系AUC来评价链接预测方法的效果。通过实验验证,该方法的预测准确率更高。

With the development of computer technology and network,data emerge explosively,and social media constantly integrate into people’s life. Social network analysis has become a research hotspot. With the advent of big data era,the research on social network link algorithm has a great impact. The original network structure-based prediction method has gradually become unsuitable for the status quo. Therefore,we propose a social network link prediction method based on topic model. Firstly,the experimental network is divided into test set and training set with the Microblog social network as the data source. Secondly,the topic model is used to obtain users’ topic features,and the similarity measure of users’ interest features is obtained by combining the named entity set and the user association feature set. Moreover,the similarity degree of user nodes is obtained by combining the network structure similarity,so as to predict the social network links.Finally,the link prediction method is evaluated by AUC,the most common evaluation system of link prediction. The experiment shows that the proposed method has higher prediction accuracy.