基于能量优化的微博用户转发行为预测

发布时间：2018-11-08 18:00

【摘要】：微博用户转发行为预测是微博社交网络消息扩散模型构建的基础,在舆情监控、市场营销与政治选举等领域有着广泛的应用.为了提高用户转发行为预测的精度,本文在MRF(Markov Random Field)能量优化框架下综合分析了用户属性与微博内容特征、用户转发行为约束、群体转发先验等因素对用户转发行为的影响,并在逻辑回归模型的基础上构造了相应的能量函数对用户转发行为进行了全局性的预测.实验结果表明,微博用户转发行为不仅取决于用户属性、微博内容等特征,而且也受到用户转发行为约束、群体转发先验等因素不同程度的影响.相对于传统算法,本文算法可以更准确地对用户转发行为进行建模,因而可获得更好的预测结果.
[Abstract]:Weibo's user forwarding behavior prediction is the basis for building a social network message diffusion model, which is widely used in the fields of public opinion monitoring, marketing and political election. In order to improve the accuracy of user forwarding behavior prediction, the user attributes and Weibo content characteristics, user forwarding behavior constraints are comprehensively analyzed in this paper under the framework of MRF (Markov Random Field) energy optimization. Based on the logical regression model, a corresponding energy function is constructed to predict the user forwarding behavior globally. The experimental results show that the user forwarding behavior of Weibo is not only dependent on the characteristics of user attributes, Weibo content, but also affected by user forwarding behavior constraints and group forwarding prior factors. Compared with the traditional algorithm, the proposed algorithm can model the user forwarding behavior more accurately, so that better prediction results can be obtained.
【作者单位】：周口师范学院网络工程学院;
【基金】：国家自然科学基金(No.U1404620,No.U1404622) 河南省自然科学基金(No.162300410347) 河南省科技攻关项目(No.172102310727,No.162102310589,No.162102210396,No.162102310590) 河南省高校重点科研项目(No.17A520018,No.17A520019,No.15A520116,No.16B520034,No.16A520105) 周口师范学院高层次人才科研启动基金(No.zknuc2015103)
【分类号】：TP393.092

【相似文献】