k-medianoids Clustering Algorithm

Abstract

One of the simplest and popular clustering method is the simple k-means clustering algorithm. One of the drawbacks of the method is its sensitivity to outliers. To overcome this problem, the k-medians clustering algorithm is used. Another limitation of the simple k-means clustering algorithm is the Euclidean space assumption. The k-medoids has been used to overcome this assumption. Here a combined method called the k-medianoids clustering algorithm is proposed. A medianoid is a kind of median that does not require the Euclidean space assumption and is formally defined. The proposed method is demonstrated using nucleotide sequences.