Efficiency of K-Prototype and K-Mean algorithm using Support Vector Machine (SVM)

Abstract

Clustering is a key method in unsupervised machine learning, which is commonly used to find latent patterns in unlabeled datasets.  This research evaluates the efficacy of K-Means and K-Prototype clustering algorithms using five benchmark datasets that include labeled, unlabeled, and mixed-type data.  After routine preprocessing, datasets were divided into 2 to 5 clusters, and a Support Vector Machine (SVM) classifier was used to check the resulting cluster assignments.  Experimental results show that K-Means works better on labeled datasets, while K-Prototype works better on unlabeled and mixed-type datasets.  Also, accuracy goes down as the number of clusters goes up, and the best results are shown with two clusters.  These results show how the type of data and the way the clusters are set up affect how well clustering and classification tasks work.