Cluster Analysis in Insulating Fluids Using $k$-Means and Hierarchical Clustering Methods
Raymon Antony Raj, Sampath Kumar Venkatachary, D Sarathkumar, J. Sivadasan, Leo John Baptist Andrews, Srinivasan Murugesan
Abstract
The research looks at the relationships that each species group has with the other groupings. Pongamia Oil, Modified-Pongamia Oil, and Mineral Oil are the three types of oil employed in the research; they are referred to by their respective species designations (here, species=category), PO, MPO, and MO. Additionally, the group is described using four dielectric parameters, including breakdown voltage, kinematic viscosity, and dielectric constant, and moisture content. These values are abbreviated as BDV, KVIS, DC, and MC in that order. Using k-Means clustering, the data are split into k mutually exclusive clusters, with centroids connected to each cluster for rearranging. The sum of distances between the centroids and other data are reduced by the k-Means algorithm using Silhouette values. These additional data are given an object identifier that may be used to subsequently reorganize the cluster. While hierarchical clustering identifies the many levels of data categorization. The cophenetic correlation is used to confirm the consistency of the cluster tree's initial distances. Dendrogram plot is also used to show the hierarchy of clusters. These clusters demonstrate that there are at least two groups for each oil, and a detailed examination of the clusters will reveal the salient characteristics of the group that will affect its qualities.