Sequential extraction of clusters for imbalanced data

K-means type clustering has a central role in various clustering algorithms. In spite of its usefulness, there is a well-known drawback, the number of clusters should be determined beforehand, and clustering results are strongly depends of this number. Many researchers study on how to estimate this number and one algorithm is using sequential extraction of clusters. However, the clustering results by this algorithm is severely affected by the initial parameter setting. Additionally, if the dataset consists of imbalanced clusters and shapes, the results also can be worse. To overcome such problems, we propose automatic estimation of parameter values during the clustering process. We show the effectiveness of the proposed algorithm by using numerical examples.