论文部分内容阅读
在电力企业同业对标管理中,K-means聚类算法作为一种无监督分类算法,无需任何先验知识就能够实现对企业的分类功能,确立标杆企业。但是,K-means算法的聚类数目k和初始聚类中心都需要人为事先给出。为了解决这些不足,提出了一种对K-means聚类改进算法:先分别通过两阶聚类法和最大距离法确定聚类数目和初始聚类中心,然后再使用K-means算法完成聚类。通过实例表明此算法可以选出优秀企业,并能发现其他企业与标杆企业之间的差距。
In the peer management of power enterprises, K-means clustering algorithm as an unsupervised classification algorithm, without any priori knowledge can realize the classification of enterprises, the establishment of benchmarking enterprises. However, both the k-means clustering number k and the initial clustering center need to be given in advance. In order to solve these deficiencies, an improved algorithm for K-means clustering is proposed. Firstly, the number of clusters and the initial cluster centers are determined by two-stage clustering and maximum distance respectively, and then K-means clustering . An example shows that this algorithm can select excellent enterprises and discover the difference between other enterprises and benchmarking enterprises.