一种中心权值流数据聚类算法

Stream Data Clustering Algorithm based on Center Weighting

近年来,不确定的聚类中心使得传统的聚类算法面临着巨大的挑战,例如导致了丢失历史数据信息、数据边界不清晰、时间复杂度高等问题。因此,提出了中心加权数据流聚类(Center-Weighted algorithm for clustering data streams,CW-Stream)算法,旨在提高聚类中心的实时性。为了保留历史数据的瞬态特征,并尽可能少地占用存储空间,聚类中心被分配了不同的权重,并使用中心权重算法代替传统的欧几里得距离来调整微簇信息,改进了模糊可扩展策略。此外,采用模糊隶属度矩阵记录动态的数据流信息。与其他经典算法相比,该算法在聚类纯度和效率方面有较好的性能。

In recent years,uncertain clustering centers have posed a great challenge to traditional clustering algorithms,such as losing historical data information,unclear boundaries and high time complexity.Therefore,a center-weighted algorithm for clustering data streams(CW-Stream)algorithm is proposed to improve the real-time performance of clustering centers.In order to preserve the transient characteristics of historical data and occupy as little storage space as possible,the cluster centers are assigned different weights,and the center weight algorithm is used instead of the traditional Euclidean distance to adjust the micro-cluster information and improve the blur Scalable strategy.In addition,a fuzzy membership matrix is used to record dynamic data flow information.Compared with other classic algorithms,this algorithm has better performance in terms of clustering purity and efficiency.