节点决策认知的在线社交信息扩散动力与有序传播机制研究

Online social networks (OSNs) are a type of social structure that comprises a set of social actors and social interactions between them. With the rapid development of the Internet and new media technologies, online social networking sites and applications, have emerged in people’s lives. Users are able to upload small pieces of information and share them with their friends in OSNs. The information would then be spread in a cascade mode by active online friends. This phenomenon motivates researchers to discover the information diffusion patterns in OSNs, and become a hot research topic. It is meaningful and challengeable to explore the dynamics of information diffusion and well-organized propagation mechanisms. The purpose of this project is to research on the online social information diffusion mechanisms and well-organized propagation methods, by analyzing the key driving effect of node decision cognition in information dissemination.. Firstly, the iterative cognitive calculation method is proposed to model and measure the key factors such as user interaction interest, customer behavior, trust relationships and interactive activities, by user information interactions cascade. . Then, the non-constant dynamic diffusion probability cascade method is proposed based on the correlation of multi-dimensional user decision-making cognition, to solve the scientific problem that traditional epidemic-like constant probability hypothesis does not full adapt to the nature of online social information diffusion.. Furthermore, inspired by physical particle dynamics, the multi-source diffusion model is proposed by merging multi-dimensional user decisions, to adapt the multi-source dissemination features in online social networks, such as multi-source competition, multi-source cooperation and multi-homologous concurrency, and to improve the propagation controllability in OSN.. Finally, on the basis of above, the link prediction algorithms based on correlation of node decision cognition is proposed to discover the latent social links. Computing methods of social influence maximization is proposed based on proposed multi-source diffusion dynamics, to discover the optimal influencers. And, the computing methods of social rumor influence minimization is proposed based on cooperative immune evolution, to limit the rumors propagation and improve the ordering of information dissemination in OSN. . All above are based on actual data feedback and optimization, and a prototype system is proposed for demonstration based on our former developed online social share system.

在线社交网络信息扩散动力学是当前研究热点，探索其信息扩散动力与有序传播机制对构建良好网络生态具有重要的科学意义与应用价值，尚存在诸多挑战。研究表明，作为节点自主连接构成的复杂网络，节点的决策对社交信息传播起到关键作用。本项目以节点决策认知为主线，系统探索信息扩散规律及其有序可控方法。首先研究节点交互兴趣、从众行为、信任关联、交互活性等节点决策关键因子度量方法，提出了行为级联的节点决策迭代认知方法。进而研究节点决策认知的社交信息激活概率模型，提出了决策关联的非恒定动态概率扩散模型，结合多源竞争、合作、并发等扩散特征，提出了决策融合的质点动力学多源扩散模型。在此基础上，研究信息扩散规律的有序可控方法，提出了决策关联的社交连接预测算法、能量级联的最优影响力节点发现算法、协同免疫的谣言传播控制方法，提高信息扩散的有序性。所做研究基于实际数据反馈优化，并在前期六维空间交互系统基础上构建原型演示系统。

线社交网络(OSN)作为一种新的内容创造和传播媒介，不仅进一步改变了人们的互动方式和互动范围，也给网络空间安全带来了挑战。本项目以探索可控的OSN信息扩散规律与有序传播方法为总体目标。本项目以探索可控的OSN信息扩散规律与有序传播方法为总体目标，以用户在信息传播中的决策认知为主线，研究OSN节点决策驱动力建模与分析方法，探索基于节点决策认知的OSN信息传播、交互预测、谣言免疫的信息扩散模型与算法，为构建有序可控的OSN生态提供技术参考。从OSN交互行为特征出发，提出了一套行为级联的节点决策建模、认知、计算模型，为用户决策的社交行为分析和预测提供了可解释的方案；基于多维决策驱动力量化分析节点决策对信息传播产生影响，提出了一套节点决策关联的动态扩散概率模型与方法；从传播的内生动力出发，借鉴物理学动力模型，提出了一套具有高近似比的融合节点决策认知的OSN多源信息扩散理论及有序传播进化机制；从扩散的应用影响出发，面向影响力最大化问题，提出了一套节点决策关联的连接预测与最优影响力节点发现算法与模型；面向恶意传播最小化问题，基于隔离阻塞提出最少免疫节点的谣言传播最小化方法，基于多源传播竞争进化提出正向竞争协同模型与算法；面向多源信息传播溯源问题提出了基于多属性拓扑聚类社区划分溯源方法。所做研究均在公开数据集上进行了正确性、先进性的验证与比对。在项目的资助下，课题组成员在国内外学术期刊和国际会议上发表标准本项目资助号的论文22篇（第一标注18篇）。其中，在Future Generation Computer Systems、Knowledge-Based Systems、IEEE Intelligent Systems、Journal of Information Security and Applications等期刊上发表SCI收录的论文14篇，在IJCAI、ICME、PCM等国际会议上发表论文6篇。对所做研究成果及时形成版权保护，获得软件著作权6件，申请“国家发明专利1件。成果获得CCF B类会议ICME的最佳论文提名奖、辽宁省自然科学学术成果一等奖。直接培养了9名硕士毕业研究生。