一种有效的K-means初始中心优化算法

局部最优是K-means算法最容易形成的一个问题,所以聚类结果会大大受初始中心的波及。。针对这一问题,找到了改进初始聚类中心的新方法:首先,选择高密度区域内距离最远的两个点为初始中心,然后将第3个初始中心位置规定在与已知初始聚类中心距离乘积最大的点上,以此类推,直到找到k个初始中心。实验证明,此算法有更快的收敛速度,生成的结果稳定性更强,正确率更高。

K -means algorithm terminates at a local optimum state, so the choice of initial centers will affect the clustering results to a largeextent. To solve the question, this paper presents a method of optimizing the initial center. The algorithm selects two points at the furthestmutual distance in high-density region as the initial cluster centers; then this paper sets up the third initial center according to the maximumdistance between the product with points found this way .Experimental results demonstrate that this method compared with the traditionalK-means algorithm has faster convergence speed and higher accuracy and greater stability.