基于特征加权理论的数据聚类算法

针对数据挖掘过程中数据聚类操作的初始聚类数目和初始聚类中心确定困难的问题,提出了一种软子空间结合竞争合并机制的模糊加权聚类算法.通过对软子空间聚类算法的目标函数进行改写,并结合数据簇势的大小对各数据簇进行竞争与合并操作,实现了对数据的聚类处理.结果表明,该算法能够准确地对数据样本进行聚类,并且聚类结果与初始数据簇数目和初始聚类中心无关,能够满足对高维数据聚类处理的需要,具有较好的实际应用价值.

Data clustering algorithm based on feature weighting theory

Aiming at the problem that the initial clustering number and center are difficult to be determined in the data clustering opertion of data mining process, a fuzzy weighting clustering algorithm based on the soft subspace as well as the competition and combination mechanism was proposed. Through rewriting the objective function of soft subspace clustering algorithm and combining the size of data clusters, the competition and combination operation was carried out for the data clusters, and the clustering treatment of data was achieved. The results show that the proposed algorithm can accurately perform the clustering of data samples, and the clustering results are independent on the initial clustering number and center. The algorithm can meet the need in high dimensional data clustering processing and has the great practical value.