Fast effective clustering technologies for highly dynamic massive networks

聚类是一项基本的数据挖掘和分析任务。在一个相互关联的不断发展的世界中，友谊和信息流被建模为大型动态网络。结构聚类法和相关性聚类法是静态网络中重要且研究较多的方法；对于不断演化的网络，其中链接随着时间的推移而出现和消失，我们缺乏有效的技术。预期的结果是新的实用的动态网络集群算法--具有效率和集群质量的性能保证--以及原型软件，指导我们挑选一个好的集群。预期的好处包括更好地了解不断发展的社交网络中的传播，加快软件测试周期，以及改进话题检测。

Clustering is a fundamental data mining and analysis task. In an interconnected evolving world, friendships and information flows are modelled as large dynamic networks. Structural clustering and correlation clustering are important and well-studied approaches for static networks; for evolving networks, where links appear and disappear over time, we lack efficient techniques. Anticipated outcomes are new practical clustering algorithms for dynamic networks – with performance guarantees of efficiency and clustering quality – and prototype software, guiding us to pick a good clustering. Expected benefits include better understanding of spread in evolving social networks, accelerating the software testing cycle, and improved topic detection.