Application of fuzzy K-means (FKM) algorithms in identifying better clusters of few drugs from drugbank database

Abstract

The fuzzification of the cluster configuration is refereed as Fuzzy K-Means (FKM) where the algorithm generates limited homogeneous clusters. The data points are assigned respective clusters in accordance to the membership degrees within interval [0,1]. Several variations of FKM algorithm were applied in identifying better clusters of few drugs data set derived from DrugBank database as possible GSK-3 beta inhibitors defined against diabetes. Better clusters were evaluated based on cluster balance and membership degree plots. With k=3, observation of cluster balance and membership degree plots revealed that FKM with entropy is the best method of choice with equal assignment of objects and no ambiguous assignments. The membership degree plot resulted in a good fuzzy clustering result where only 3 points appeared between membership degrees of 0.6 to 0.8.