[论文解读] A Method for Group Extraction and Analysis in Multilayer Social Networks
本文提出了一种新颖的方法,用于在多层社交网络中进行群体提取与演化分析,采用CLECC度量——一种跨层边聚类系数,用于量化跨层邻居的相似性。CLECC算法用于识别社区,而群体演化发现(GED)方法则追踪随时间的群体变化,在真实网络和基准网络上实现了高准确率与高效率,包括为多层网络新设计的mLFR基准。
The main subject studied in this dissertation is a multi-layered social network (MSN) and its analysis. One of the crucial problems in multi-layered social network analysis is community extraction. To cope with this problem the CLECC measure (Cross Layered Edge Clustering Coefficient) was proposed in the thesis. It is an edge measure which expresses how much the neighbors of two given users are similar each other. Based on this measure the CLECC algorithm for community extraction in the multi-layered social networks was designed. The algorithm was tested on the real single-layered social networks (SSN) and multi-layered social networks (MSN), as well as on benchmark networks from GN Benchmark (SSN), LFR Benchmark (SSN) and mLFR Benchmark (MSN) a special extension of LFR Benchmark, designed as a part of this thesis, which is able to produce multi-layered benchmark networks. The second research problem considered in the thesis was group evolution discovery. Studies on this problem have led to the development of the inclusion measure and the Group Evolution Discovery (GED) method, which is designed to identify events between two groups in successive time frames in the social network. The method was tested on a real social network and compared with two well-known algorithms regarding accuracy, execution time, flexibility and ease of implementation. Finally, a new approach to prediction of group evolution in the social network was developed. The new approach involves usage of the outputs of the GED method. It is shown, that using even a simple sequence, which consists of several preceding groups sizes and events, as an input for the classifier, the learned model can produce very good results also for simple classifiers.
研究动机与目标
- 解决多层社交网络(MSNs)中的社区检测挑战,传统单层方法因结构复杂性而失效。
- 提出一种新的边级别度量CLECC,用于评估多层网络中邻居的相似性,从而实现稳健的社区检测。
- 通过识别动态社交网络中结构变化(如合并、分裂、扩展)来实现群体随时间演化的发现。
- 创建一个新的多层基准(mLFR),用于在真实、可扩展的多层网络结构上评估算法。
- 设计一个基于历史群体规模和事件序列的群体演化预测框架,即使使用简单分类器也能实现优异性能。
提出的方法
- 提出CLECC(跨层边聚类系数),一种用于评估两个节点在多层网络中共同邻居相似性的度量。
- 将CLECC用作社区检测算法中的加权机制,通过聚合边的相似性形成紧密群体。
- 开发群体演化发现(GED)方法,利用一种新颖的包含度量来检测连续时间帧之间的群体转换(如合并、分裂)。
- 实现一个预测模型,将先前的群体规模序列和演化事件作为输入特征,用于分类未来群体状态。
- 构建mLFR基准,作为LFR基准的扩展,用于生成具有受控社区结构和层特定连通性的真实多层网络。
- 使用标准化指标(如标准化互信息和调整兰德指数)在真实世界单层与多层网络以及新创建的mLFR基准上,评估CLECC与GED方法。
实验结果
研究问题
- RQ1在交互跨越多种关系类型的情况下,如何在多层社交网络中有效进行社区检测?
- RQ2与单层方法相比,CLECC度量在多层设置下能在多大程度上提升社区检测的准确率?
- RQ3在动态多层网络中,群体演化事件(如合并、分裂)随时间检测的准确度如何?
- RQ4仅使用历史群体规模和事件序列的简单分类器,能否以高准确率预测未来的群体演化?
- RQ5所提出的方法在性能、执行时间与灵活性方面,与现有最先进的社交网络群体分析算法相比如何?
主要发现
- CLECC算法在真实和基准多层网络中均实现了高准确率的社区检测,优于单层基线方法。
- 与两种知名算法相比,GED方法在检测群体演化事件方面表现出更优的准确率与效率,执行时间更短,实现更简便。
- 仅使用先前群体规模和演化事件序列的预测模型即实现了优异性能,表明简单的时间序列对预测群体动态具有高度信息量。
- 新开发的mLFR基准为多层社区检测与演化方法的可靠评估提供了支持,为未来研究提供了标准化测试平台。
- CLECC与GED的集成构建了从社区提取到演化追踪与预测的完整分析流程。
- 实证评估证实,所提出的方法具有灵活性、可扩展性,适用于真实世界社交网络分析应用。
更好的研究,从现在开始
从阅读论文到最终审阅,大幅缩短您的研究时间。
无需绑定信用卡
本解读由 AI 生成,并经人工编辑审核。