AK-means: an automatic clustering algorithm based on K-means

In data mining, K-means is a simple and fast algorithm for solving clustering problems, but it requires that the user provides in advance the exact number of clusters (k), which is often not obvious. Thus, this paper intends to overcome this problem by proposing a parameter-free algorithm for automatic clustering. It is based on successive adequate restarting of K-means algorithm. Experiments conducted on several standard data sets demonstrate that the proposed approach is effective and outperforms the related well known algorithm G-means, in terms of clustering accuracy and estimation of the correct number of clusters.