摘要
影响力最大化是近年来广泛研究的社交网络的核心问题.然而之前的研究较少考虑用户的意愿以及用户之间的友好或敌对关系.因此综合考虑这些因素,针对符号网络提出了考虑用户意愿的净积极影响力最大化问题,该问题可以描述如下:利用符号网络来刻画用户具有友好(积极)和敌对(消极)关系的社交网络,每个用户对传播的信息有自己的意愿,目标是要从网络中选择k个用户,使得最终的净积极影响的用户数量最多.通过对问题的细致分析,建立了考虑用户意愿的传播模型,证明了该模型下净积极影响力最大化问题是非次模和非单调的,随后给出了基于概率驱动的结构感知的求解算法,通过在三个数据集上的实验表明,利用提出的算法找到的种子集有更好的净积极影响力.
The influence maximization problem was a critical problem in social networks and had been extensively studied.Previous studies had given less consideration to user’s willingness and friendly or hostile relationships between users.Therefore,considering these factors,the problem of maximizing the net positive influence of the signed network considering the user's willingness was proposed.The problem could be described as follows:using a signed social network to portray a social network with friendly(positive)and hostile(negative)relationships between users,each user had his own willingness for the information spreading,and the goal was to select k users from the network,so that the net positive influence on users was the most.Through the detailed analysis of the problem,proposed a propagation model,proved that the problem was neither monotonic nor submodular under the model.At last proposed a algorithm based on probability driven structure-aware algorithm.Experiments on three datasets showed that the seed set found using the proposed algorithm had a better net positive influence.
作者
宗金醒
帅天平
ZONG Jinxing;SHUAI Tianping(School of Science,Beijing University of Posts and Telecommunications,Beijing 100876,China)
出处
《哈尔滨商业大学学报（自然科学版）》
CAS
2023年第5期604-611,共8页
Journal of Harbin University of Commerce：Natural Sciences Edition
基金
国家自然科学基金项目(No.12171051,12171052)
中央高校基本科研业务费(No.500421358)
北京邮电大学提升科技创新能力行动计划项目(No.2020XD-A01-1)。
关键词
符号网络
积极影响
消极影响
净积极影响力
概率驱动结构感知算法
用户意愿
signed social network
positive influence
negative influence
net positive influence maximization problem
probability driven structure-aware algorithm
user’s willingness
