A Comparative Analysis of Clustering Methods on the 20 Newsgroups Dataset for Analytics

This paper presents a comparative analysis of two different approaches for clustering textual data from the 20 Newsgroups dataset. The first approach leverages a Large Language Model (LLM) to classify each text into predefined categories using zero-shot classification. The second approach applies to the traditional K-Means clustering algorithm on text embeddings. We evaluate both methods by comparing their predicted clusters against true labels for assessment. For K-Means, we also explore a semi-supervised variant with centroid initialization based on true labels.