论文部分内容阅读
随着社交网络的发展,用户影响力研究得到广泛关注。Page Rank作为分析用户影响力的传统算法,计算用户影响力时只考虑了用户的粉丝数和关注数,并没有考虑用户本身的行为对影响力传播造成的影响。文章综合考虑了用户行为中的转发、评论、艾特数,论证了随着时间的变化用户行为符合幂率分布,以此定义了兴趣度模型,用来研究用户行为随时间的变化规律。充分将用户个体的行为融入到传统的影响力评估模型中,定义了活跃度的概念,可剔除不活跃的用户。此外,摒弃了Page Rank中将影响力权值平均分配的方法,重新定义了影响力分配因子,对于信息传播过程中影响力贡献较大的用户给予更高的分配比例,使得影响力的评估更加准确。实验结果表明,本文提出的基于用户行为的实时影响力算法(User Real-time Influence),在提高影响力计算的准确性,更加符合社交网络的特性。
With the development of social networks, research on user influence has drawn wide attention. As a traditional algorithm for analyzing user influence, Page Rank only considers the number of fans and the number of users when calculating user influence, and does not consider the effect of users’ own behaviors on impact communication. In this paper, we consider the forwarding, commenting and the number of Atites in the user behavior, and demonstrate that the user behavior complies with the power distribution over time, so as to define the interest degree model, which is used to study the change rule of user behavior with time. Fully integrate the individual behavior of users into the traditional impact assessment model, define the concept of activity, can eliminate inactive users. In addition, we abandon the method of average distribution of influence weight in Page Rank, redefine the influence distribution factor, give a higher allocation proportion to those who contribute more influence in the process of information dissemination, and make the assessment of influence more accurate. The experimental results show that the proposed User Real-time Influence based on user behavior improves the accuracy of influence calculation more in line with the characteristics of social networks.