基于点概率的K-means算法的改进

Improved K-means Algorithm Based on Dot-probability

K-means算法是一种基于划分的方法,该算法对初始聚类中心的选取依赖性极大,初始中心值的不同导致聚类效果不稳定.为此,本文利用几何概率的思想,认为每个数据点都是等概率的存在于数据集,通过计算每个数据点的点概率值,结合距离因素,选择K个点作为初始聚类中心.实验证明,改进后的K-means算法聚类效果更好.

K-means algorithm is a division-based method,which is greatly dependent on the choosing of initial cluster centers. Dif-ferent initial clustering center value can lead to unstable destabilizing effect. Thus,this article holds the idea that each data point in the da-ta set has the same probability through calculating dot-probability value for each data point and combining with the distance factor tochoose K points as the initial cluster centers by using the principle of geometric probability. Experiment shows that the improved K-meansalgorithm clustering effect is better.