基于信任结构平衡的社交网络用户群组演化计算与分析关键技术研究

The procedure computation of user group evolution is a significant research field in social network analysis. However, there are several limitations in existing researches of social network groups, including mining groups based on heterogeneous features and neglecting describing the internal causes of group construction in a unified way, discovering groups from static perspective and ignoring its evolution process, and lacking the group evolution computational theories with combining both the logicality of graph computation and the intelligence of deep learning. The main goal of our project focuses on analyzing the foundations of social network group evolution process from a dynamic perspective. Therefore, the key scientific problems of social network group evolution of this project are refined as: giving a new logical description basis and proposing a new process computation methodology for user group evolution. And further, we propose the research ideas purposefully as: proposing the trust structure model for describing the internal cause of group evolution and presenting the trust structure balance computation for supporting the external phenomena computation of group evolution. The main works of this project are as: firstly, the trust structure model is given as a logical description basis to characterize the unified internal cause that promotes the processes of group construction and evolution; secondly, the description method of social network groups are proposed based on trust structure model so that the construction of group is supported by the proposed internal cause; thirdly, trust structure balance judgment method is addressed for expressing the status changing for group evolution, which is given based on the deep learning method to support social network big data intelligent analysis; and finally, the group evolution process computation method is presented based on the proposed trust structure balance judgment, and then the group evolution lifecycle model is established for describing the whole group evolution phrases in social network. From this project, we try hard to provide novel ideas and benefits for social network group mining and analysis.

用户群组演化过程计算是社交网络研究的重要领域。然而，当前社交网络群体分析研究局限在于：以差异化特征挖掘群体而忽视群体构成统一抽象内因的描述、以静态群体发现而忽视其动态演化过程，以及缺乏兼顾图计算逻辑性与深度学习智能性的演化计算理论。本项目以动态视角实现社交网络群组演化全过程计算与分析为目标，凝练出研究的关键科学问题：群组演化过程的“逻辑描述新基础”和“过程计算新方法”，针对性地提出“以信任结构为群组演化驱动内因，以信任结构平衡计算支撑群组演化过程计算”的思路：提出以信任结构模型描述为群组演化统一内因提供逻辑基础；提出基于信任结构的社交网络群组描述方法，使群组构成得到内因支撑；以深度学习支持大数据智能分析，给出信任结构平衡判定方法，使群组演化内因变化具备计算基础；最后，提出基于信任结构平衡的群组演化过程计算，建立群组演化生命周期模型。项目力图为社交网络群体挖掘与分析研究提供全新的思路。

社交网络用户群组反映了群体状态及其发展演化的趋势。然而，当前社交网络群体分析研究局限在于：以差异化特征挖掘群体而忽视群体构成统一抽象内因的描述、以静态群体发现而忽视其动态演化过程，以及缺乏兼顾图计算逻辑性与深度学习智能性的演化计算理论。本项目研究并提出了动态视角实现社交网络群组演化全过程分析的相关方法与模型，致力于实现社交网络群组状态的过程性计算与结果状态预测的融合。项目聚焦于两个关键科学问题研究：群组演化过程的“逻辑描述新基础”和“过程计算新方法”，主要完成了三个方面的研究工作： 首先，项目提出以信任结构为群组演化驱动内因，通过群体内信任关系的结构平衡状态定义与计算，判定用户群组本身形成并得以维持的内部驱动力；其次，基于信任结构的形式化描述定义，提出了用户信任群组构建的基本方法；第三，以深度学习支持大数据智能分析，通过结构特征的分类挖掘，给出了信任结构平衡判定方法，使群组演化内因变化具备了可量化计算的基础；最后，通过结构平衡特征的时序变化趋势，提出了基于信任结构平衡的群组演化过程计算模型，建立群组演化生命周期模型。项目以实际社交网络微博为实证数据对象，开展了相应的数据分析研究。本项目已发表相关论文16篇，其中SCI/EI检索共计15篇；申请发明专利9项；获得软件著作权9项；项目研究的网络大数据分析技术、结构特征挖掘技术等核心技术成果，同时获得了多项奖励。