复杂社交网络中基于Agent模型的个体观点预测

发布时间：2018-01-14 05:04

本文关键词：复杂社交网络中基于Agent模型的个体观点预测　出处：《东南大学》2017年硕士论文　论文类型：学位论文

【摘要】：近年来,随着移动终端产品的普及和社交平台的不断发展,人们越来越习惯使用社交网络平台进行人与人之间的沟通交流。然而,社交平台的匿名化慢慢滋生了一群键盘侠,同时,用户的从众心理及法不责众的心态也使网络极端事件频有发生,且由于社交平台中的信息扩散速度之快,政府等相关部门不易做好防范措施。针对上述问题,就复杂社交网络中的个体对某一话题的观点进行预测,对维护网络社会的和谐和安全是充分且必要的。众所周知,不同的个体对同一话题会持有不同的观点,且个体对他人观点的认可程度会略有差异,受他人观点的影响程度也会不尽相同。因此,为较好反映个体在形成自我观点时考虑的各类影响因素,本文综合考虑复杂自适应系统中的Agent概念,提出并设计Agent个体模型和Multi-Agent观点交互模型,从微观角度分析社交网络中某一话题社区观点演变过程,并能预测个体对该话题产生的观点倾向。为达到较好的预测效果,本文首先从抓取的实证数据中挖掘出若干话题社区,建立Multi-Agent话题社区环境,并在此基础上对社区中所有博文及评论文本信息隐含的情感倾向进行挖掘。建立话题社区网络动态模型,模拟社交网络节点间的关注关系及交互关系。其次,建立基于Agent的个体模型,设计Agent结构,将个体接收信息产生观点的过程映射为Agent通过感知器、控制器、效应器处理消息生成观点的过程,使话题社区中的Agent具有自主性、适应性、社交性、交互性四大特点。并将个体自信度、邻居依赖度、社区趋同度三个特征作为Agent个体形成观点过程中的影响因素。随后,建立Multi-Agent观点交互模型,实现话题社区环境中Agent间通讯、交互。并设计Agent个体观点交互规则生成算法,学习特征权值向量。本文在新浪微博实证数据上对Agent个体模型和Multi-Agent观点交互模型的有效性进行了验证。实验结果表明:(a)从微观个体角度分析,基于Agent的个体模型和基于Multi-Agent观点交互模型能够较准确的预测个体观点变化的转折点,对话题社区活跃个体的预测结果与实际结果出入甚微。(b)从宏观话题社区角度分析,利用基于Agent的个体模型和基于Multi-Agent观点交互模型能够较好预测话题社区所有个体的观点变化趋势,且可预测极端群体的规模。
[Abstract]:In recent years, with the popularity of mobile terminal products and the continuous development of social platform, people are more and more used to use social network platform to communicate between people. The anonymity of social platform has slowly bred a group of keyboard knights, at the same time, the mentality of users' conformity and blame also makes the network extreme events occur frequently, and because of the rapid spread of information in the social platform. The government and other relevant departments are not easy to take precautions. In response to the above problems, the individuals in the complex social networks on a certain topic point of view are predicted. It is necessary and sufficient to maintain the harmony and security of the network society. As we all know, different individuals will hold different views on the same topic, and the degree of recognition of the views of others will be slightly different. Therefore, in order to better reflect the various factors that individuals consider in the formation of self-view, this paper comprehensively considers the concept of Agent in complex adaptive systems. This paper proposes and designs the Agent individual model and the Multi-Agent viewpoint interactive model, and analyzes the evolution process of a topic community viewpoint in social network from the micro point of view. And can predict the individual opinion tendency on this topic. In order to achieve better prediction effect, this paper first excavates a number of topic communities from the collected empirical data. The Multi-Agent topic community environment is established, and on this basis, the emotional tendency of all blog posts and comments text information in the community is excavated, and the dynamic model of topic community network is established. Simulation of social network nodes between the relationship of concern and interaction. Secondly, establish an individual model based on Agent, design the structure of Agent. The process of individual receiving information to produce ideas is mapped to the process of Agent processing message generation ideas through perceptron controller and effector so that Agent in the topic community has autonomy and adaptability. Social, interactive four characteristics. And the individual confidence degree, neighborhood dependence, community convergence degree of three characteristics as the Agent individual formation of the process of influence factors. Establish Multi-Agent viewpoint interaction model, realize communication and interaction between Agent in topic community environment, and design Agent individual viewpoint interaction rule generation algorithm. The effectiveness of the Agent individual model and the Multi-Agent viewpoint interaction model is verified based on the Sina Weibo empirical data. The experimental results show that the Agent individual model and the Multi-Agent viewpoint interaction model are effective. From a microscopic perspective. The individual model based on Agent and the interactive model based on Multi-Agent viewpoint can accurately predict the turning point of individual viewpoint change. The prediction results of active individuals in the topic community are only slightly different from the actual results. (B) from the perspective of macro-topic community analysis. The individual model based on Agent and the interactive model based on Multi-Agent viewpoint can predict the trend of change of opinion of all individuals in the topic community, and can predict the scale of extreme group.
【学位授予单位】：东南大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.09

【相似文献】