Abstract
This paper presents a cluster-based analysis of political perception among social media users based on survey data collected during a pre-election period. Using K-Means clustering and PCA-based visualization, we identified four distinct user profiles that differ in their levels of exposure to political content, platform preferences, trust in information, and engagement behavior. The resulting clusters-Passive Mainstream Users, Politically Exposed Observers, Engaged Sharers, and Disconnected Users-reveal significant variation in how individuals interact with political information in digital environments. Our findings challenge the assumption that high-frequency social media use correlates with political involvement and suggest that content format, trust, and platform algorithms play a crucial role in shaping political awareness. The study contributes to the understanding of data-driven profiling in political communication and demonstrates how machine learning methods can enhance the analysis of sociological survey data without relying on content analysis.