对k-means聚类算法的改进

Improved k-means Clustering Algorithm

提出了一种k-means聚类算法中寻找初始聚类中心的新方法。算法首先计算样本间的距离,然后根据样本点之间的距离寻找有可能是一类的数据,依据这些样本点形成初始聚类中心,从而得到较好的聚类结果。实验表明,改进后的方法相对于随机选取初始聚类中心具有较高的准确率。

This paper investigates the standard k-means clustering algorithm and gives an improved algorithm by selecting better initial centers that the algorithm begins with.First the paper computes distances between data points;then tries to find out the data points that are similar;finally constructs initial centers according to these data points.In the experiment,authors find that different data points lead to different results.If people can find initial centers that are consistent with the distribution of data,people could get good clusterings.According to the experiment,the improved k-means Clustering Algorithm can get higher accuracy.