基于INF-Deffuant 模型的网络用户观点演化分析

Evolutional Analysis of Network Users’Opinion Based on an INF-Deffuant Model

复杂网络中对用户观点进行演化分析能够挖掘用户的态度变化。当前主流方法过分依赖观点相似相交假设,忽略了影响力在用户观点交互过程中的控制作用,观点预测精度不高。本文在Deffuant模型框架内引入影响力参数,提出了一种基于影响力的用户观点交互模型(INF-Deffuant)。模型原理是:当两个个体相遇时,除了观点差值较小的个体之间容易发生交互之外,个体影响力差值较大的个体之间也容易发生交互;通过运用影响力和舆论环境构造个体观点接收系数,实现了个体之间观点的动态吸收。在实验数据集上验证了INF-Deffuant模型,并与其它基于Deffuant的模型进行了对比。实验结果表明:本文方法在群体观点演化预测和用户实体观点演化预测两个层面都达到了良好的效果。

The opinion evolutional analysis in complex networks can mine the changes of users'attitudes.The current mainstream methods rely too much on the assumption of similar intersection of views,and ignore the control role of influence in the process of the opinion interaction,which leads to the low accuracy of opinion prediction.we propose an INF-Deffuant model by introducing the influence parameter into the Deffuant-based model.The principle is that when two individuals meet,in addition to the interaction between individuals with small opinion difference,individuals with large individual influence difference are also easy to interact.By using influence and public opinion environment to construct individual opinion acceptance coefficient,the dynamic absorption of opinion among individuals is realized.Our approach is verified on the experimental datasets,and compared with other Deffuant-based models.The experimental results show that this method achieves better performance in both group view evolution prediction and user entity view evolution prediction.