基于加权联合矩阵的演化聚类算法

Weighted co-association matrix oriented evolutionary clustering

传统的演化聚类算法大多是基于单个时间截面进行问题求解,对于多时间截面的融合问题尚无有效的处理办法,造成了大量的知识浪费。从时间平滑框架出发,借鉴组合聚类思想,提出一种基于加权联合矩阵的演化聚类算法(WCEC)。实验表明,该方法不仅简单有效,而且对于数据点变化的演化情况具有较高的扩展性。

Compared with static clustering, evolutionary clustering can not only resist to the noise in short-tetm, but also reflect the changing trend in long term. It has been widely used in dynamic community identification, financial product analysis and many other fields. Traditional evolutionary clustering focus on a single time step, while falls short of dealing with multiple ones. Based on the time smoothing framework, this paper put forward a weighted co-association matrix oriented evolutionary clustering（WCEC） , which proved to be simple as well as scalable through experiments.