融合用户内容与关系结构的用户影响力算法

为快速检测出信息传播的途径,减少恶意信息造成的影响,提出了一种迭代的融合用户内容与关系结构的用户影响力算法(CSIAI)。该算法通过用户微博内容建模,迭代计算出词-用户文档的相似性;另外通过微博的关注和被关注行为,建立用户关系结构,计算用户影响力权值,得到用户的影响力邻接矩阵,提取k个较大影响力的节点作为信息传播的路径。在检测仿真实验中,CSIAI以影响覆盖率和响应时间作为评价指标,根据扩充后的新知识库,确定CSIAI中参数α和β的关系。随着用户数量增长,CSIAI的影响覆盖率和响应时间性能明显优于PageRank、CELF和非迭代的融合用户内容与关系结构的用户影响力算法(CSIA)。实验结果表明,CSIAI能有效地检测到信息的传播情况。

User influence algorithm based on user content and relational structure

In order to rapidly detect the information dissemination ways and alleviate the influence of malicious information, a user Content and Structure-based Influence Algorithm with Iteration (CSIAI) was proposed. The word-user documentation similarity was iteratively computed by the proposed algorithm through the content modeling of user's microblog. Through the concern and attention behaviors of microblog, user relational structures were established and user influence weights were calculated to get the adjacency matrix of user influence. The k nodes with higher influence were extracted as the information transmission path. In the detection simulation experiments, the influence coverage rate and response time were adopted as the evaluation indexes, According to the expansion of the new knowledge base, the relationships of parameters α and β of CSIAI were determined based on the extended new knowledge base. With the increase of users, the influence coverage rate and response time performance of the proposed CSIAI are superior to the algorithms of PageRank, CELF and Content and Structure-based Influence Algorithm (CSIA) without iteration. The experimental results show that the proposed CSIAI can effectively detect the dissemination of microblog information.