Investigation of Clustering Methods for SDSS Galaxy Images through Feature Extraction with VGG-16

In cosmology, Galaxy clusters provide crucial information about the evolution of Universe. Hence identifying galaxy clusters is essential for in-depth understanding of the Universe. This article evaluates performance of four clustering methods on images from the Sloan Digital Sky Survey (SDSS). Initially features are extracted using VGG-16, then PCA is applied before applying to various clustering methods. This study evaluates the performance of K-means, hierarchical clustering, DBSCAN and Gaussian mixture model and using silhouette score, Davies-Bouldin index, and visual inspection, we assess the quality of each clustering approach. Experimental results show that except DBSCAN other algorithms perform satisfactorily.