Performance Optimization of Document Clustering for Harry Potter Series Comments using Cosine Similarity

Abstract

This research delves into the distinctive realm of comment clustering, focusing on the extensive discourse generated by the Harry Potter series. Leveraging a dataset from Kaggle, the study aims to optimize document clustering using cosine similarity within the K-Means algorithm. The research addresses the nuanced dynamics of sentiment and preferences within the Harry Potter fan community. A comprehensive methodology involves data collection, preprocessing, TF-IDF initialization, K-Means clustering with varying distance metrics, and result evaluation. The dataset of 491 respondents unveils diverse gender, geographical, and age distributions, adding complexity to the analysis. The K-Means clustering results highlight predominant positive sentiment, emphasizing the enduring popularity of the series. The study's originality lies in its focus on the Harry Potter cultural phenomenon, contributing to sentiment analysis and fan engagement discourse. The implications extend to researchers, practitioners, and enthusiasts seeking a deeper understanding of online discussions surrounding iconic media franchises.