K-Means Clustering Algorithm with Refined Initial Center
Xuhui Chen, Yong Xu · 2009
K-means algorithm is a popular method in clustering analysis. After reviewing the traditional K-means algorithm, we proposed an improved K-means algorithm. At first we select the Euclidean distance or Manhattan distance as distance measure in our algorithm through calculating the rule of distance measure. Different initial centroids lead to different results. So the next step we will select the initial centroids which are consistent with the distribution of data. According to simulation, the improved K-means algorithm has can achieve higher accuracy and stability than the traditional ones.