基于PageRank的用户影响力评价改进算法

Improved user influence evaluation algorithm based on PageRank

摘要:

为了解决传统微博用户影响力评价算法全面性和客观性差的问题，通过对微博用户影响力的定义和影响因素进行分析，鉴于微博社区网络与web页面网络的拓扑结构有着天然相似性的特点，提出了一种基于PageRank的用户影响力评价改进算法(Self and Followers User Influence Rank)SF-UIR.运用用户追随者数、用户是否认证、用户微博的传播能力三个指标对用户自身影响因素进行了量化，改善了PageRank值对用户影响力评价客观性差的问题.采用权重因子将追随者对其所关注用户的影响力贡献值进行科学的量化分配，解决了追随者影响力等值传递的弊端.与四类主流算法的对比实验结果表明：SF-UIR算法同时考虑了基于用户行为的自身影响因素和基于拓扑结构的追随者影响因素，能够有效地解决追随者数量排名算法中的“僵尸粉”干扰问题，能比平均转发数算法更真实地反映用户的影响力高低，能有效规避K-覆盖度算法中未考虑微博用户自身行为特征和将所有的追随者都一视同仁的严重缺陷，能极大地改进PageRank算法单纯依赖追随者数量和追随者质量的不足，从而能够更加全面、更加客观地反映微博用户的影响力.

Abstract:

To solve the less comprehensive and objective problem of the traditional microblog user influence evaluation algorithms, through the analysis of the definition and influencing factors of microblog user influence, this paper proposes an improved user influence ranking algorithm based on PageRank algorithm, named as Self and Followers User Influence Rank (SF-UIR). The user's own factors are quantified by using the three indicators, the number of followers, the situation of certification, and the microblog dissemination ability, and the poor objectivity situation of PageRank values for user influence ranking is improved. The disadvantage of influence equivalent transfer of the followers' influence is overcame by adopting weighting factor to distribute the influence contribution value of different followers scientifically and quantitatively. Compared with the four mainstream algorithms, the results show that the proposed algorithm is more comprehensive, more objective, and can reflect the influence of microblog users better because of considering the influencing factors based on the user's behavior and followers factors based on the topology, which can effectively solve the interference problem of "zombie fan" in a number of followers ranking algorithm. It can reflect the user's influence level more realistically than average forwarding number algorithm, and can availably avoid the serious defects of not taking the microblog user's behavior into account and giving equal treatment to all followers in K-coverage algorithm. The proposed algorithm can greatly improve the shortage of relying solely on the quantity and quality of followers in PageRank algorithm.