Identification of Best Cluster from a Datasets Using Hybrid Hierarchical and Non-Hierarchical Methods

A hybrid hierarchical and non-hierarchical clustering approach with a prominent presence in eating disorders has been proposed here for a group of 52 licenced marketed medications proven to act as an inhibitor of 5- HT receptors. In literature, various clustering algorithms have been put forward and k-means and hierarchy-clustering algorithms are widely used. A novel approach hybrid hierarchical k-means (hk means) algorithm was proposed with the option to extract the optimum number of clusters from critical data. The statistics of Hopkins (H) is calculated to evaluate whether it could be clustered before the study. Using the statistic of Hopkins, 0.2357 was found that data are strongly clusterable to 5T receptor drug compounds. Using the consensus method, which involves about 30 NbClust indices was used in the latter to decide the optimum number of clusters. Based on NbClust’s optimum clusters, hk is an algorithm that first generates 3 cluster centres, which are fed k-means of the hierarchical clustering algorithm.