基于约束满足的大数据聚类中心点确定仿真

Big Data Clustering Center Point Determination Simulation Based on Constraint Satisfaction

针对传统的大数据聚类中心点确定方法存在用时较长、准确性较低等问题,提出了一种基于约束满足的大数据聚类中心点确定方法。将数据分布密度与增加数据关键点密度权值两种方法相结合,对大数据初始聚类中心进行K-means聚类,并获取最优聚类数目。通过最优聚类数目构建微型相似性矩阵,采用Gabow算法提取该矩阵所对应连通图的各个强连通分支。在强连通分支的基础之上,通过约束传播算法获取整个数据集的点对相似度,并利用点对相似度和奇异值分解确定大数据聚类中心点,实现数据聚类。实验结果表明,所提方法对具有更高的聚类准确性以及更低的聚类时间,适合海量数据的聚类应用。

The traditional method for determining the center point of big data clustering is time consuming.In this article,a method for determining the center points of big data clustering based on constraint satisfaction was put forward.The method of data distribution density was combined with the method of increasing the density weight of data key points to perform K-means clustering on the initial clustering centers of big data,so that the best number of clusters was obtained.Then,micro-similarity matrix was constructed by the best number of clusters.Gabow algorithm was used to extract strong components of the connected graph corresponding to this matrix.Based on strong components,the similarity of dotted pairs of whole data set was obtained by constraint propagation algorithm.Meanwhile,the center points of big data clustering were determined by similarity of dotted pairs and singular value decomposition,and thus to achieve the data clustering.Simulation results show that the proposed method has higher clustering accuracy and lower clustering time.Thus,this method is suitable for clustering application of massive data.