Optimal Value for Number of Clusters in a Dataset for Clustering Algorithm

Abstract

It is essential to know the parameters required to clustering the dataset. One of the parameters is the number of clusters k and it is very important to select the k value to get deficient results on clustering. There are few algorithms to find the k value for k-means algorithm and it requires specifying a maximum value for k or a range of values for k as an input. This paper proposes a novel method Optimal cluster number estimation algorithm (OCNE) to find the optimal number of clusters without specifying the maximum or range of k values or knee point detection in the graph. In the experiment, this method is compared with the different existing methods with deficient real-world as well as synthetic datasets and provides good performance.