User Community Discovery through Clustering of Movie Rating Patterns

Abstract

This thesis explores the discovery of user communities based on movie rating behavior using clustering techniques. By analyzing patterns in the MovieLens dataset, the study identifies groups of users with similar preferences and rating tendencies. Differences between clusters are examined in terms of engagement, rating behavior, and genre preferences. Agglomerative hierarchical clustering is applied to a user–movie rating matrix to uncover meaningful user segments. The results reveal distinct groups, including highly active users and more typical users, as well as variations between generous and critical rating behaviors. Statistical analysis confirms that these differences are significant.The findings suggest that clustering can capture meaningful behavioral patterns despite data sparsity. Additionally, cluster-based recommendations show potential for improving personalization compared to global recommendation approaches.