基于活动网络的微博用户影响力分析

Analysis of User Influence in Microblog Based on Activity Network

传统对微博用户影响力的预测仅通过微博网络的静态拓扑作结构为分析基础,忽略了用户之间的真实交互情况或者仅利用了交互行为在单个目标上的聚合结果,没有考虑用户间交互行为的强弱和相互作用等问题。对此,设计并提出一种基于活动网络的微博用户影响力分析方法—ANR。ANR利用用户之间的交互行为构建微博用户活动网络,以活动网络为基础,考虑用户间的交互强度以及用户在网络中的活跃度,对用户属性特征和交互行为特征进行了量化,进而利用在该模型上的改进PageRank算法对用户影响力进行预测。实验结果表明,ANR算法的预测结果比基于粉丝数排序的方法和基于好友关系网络的PageRank方法的预测结果更加准确和合理。

The traditional ways of predicting the impacts of microblog users are through microblogging network static topology as the basisof analyzing,ignoring the real interaction between users or only using aggregation results of a single target without considering the interac-tions and their strength between users. For this,we propose a new method for analysis of user influence based on activity network,calledANR. It firstly uses the interaction behaviors between users to construct activity network model,based on which we quantize feature ofindividual attribute and interaction behavior considering the strength of interaction between the users and user activity,thus applying theimproved PageRank algorithm to predict the user influence. The experiment shows that the ANR is more accurate and rational than thesorting method based on the number of fans and PageRank based on friend relationship network.