基于划分的K-均值初始聚类中心优化算法

K-means Initial Clustering Center Optimal Algorithm Based on Partitioning

提出了一种新的初始化K-means的聚类算法,该算法通过区域划分方法估算出K个中心点作为初始聚类中心,从初始聚类中心出发,应用K-means聚类算法,得到聚类结果.实验表明,该算法能产生高质量的聚类结果、较少的迭代次数,优于K-means算法中传统的聚类中心初始化算法.

A new method of finding the initial center of K-means algorithm is given. It first estimates the cluster center; and then initializes the K-means algorithm with the estimated center. The experiment demonstrates that the new method can produce high quality cluster result and is superior to the initialization method of cluster centers.