Article

Analyzing Patterns of Political Polarization in Contemporary Digital Society through Unsupervised Machine Learning and Data-Driven Clustering 

---

Abstract

This study investigates the spatial and algorithmic dimensions of political polarization in the contemporary digital society by analyzing U.S. county-level election data through unsupervised machine learning. Using K-Means clustering combined with Principal Component Analysis (PCA), the research identifies four distinct voting clusters that represent structural variations in partisan alignment across the United States. The Elbow Method determined that k = 4 offers the optimal balance between model simplicity and interpretability, while the PCA results indicate that the first two principal components explain approximately 34.97% of the total variance in the dataset. The clustering outcomes reveal that counties exhibit strong spatial coherence in political behavior, forming ideologically homogeneous regional blocs that reflect persistent partisan divisions. The normalized voting heatmap further demonstrates that polarization is predominantly organized along the Republican (REP) and Democratic (DEM) axis, with minor parties contributing minimally to aggregate voting patterns. These findings suggest that digital communication technologies amplify existing regional ideological identities through algorithmic filtering and selective exposure, thereby reinforcing the two-party dominance rather than diversifying political participation. The study concludes that polarization in the digital age is not only a matter of ideological difference but also a function of spatial and algorithmic reinforcement, where online and offline environments jointly shape political homogeneity. The results underscore the need for inclusive digital infrastructures, improved media literacy, and algorithmic transparency to mitigate polarization and promote more deliberative democratic engagement in the digital public sphere.

Keywords: Political Polarization, Digital Society, Machine Learning, Algorithmic Reinforcement, Spatial Clustering

How to Cite:

Budiyanto, H., (2026) “Analyzing Patterns of Political Polarization in Contemporary Digital Society through Unsupervised Machine Learning and Data-Driven Clustering ”, Journal of Digital Society 2(1). doi: https://doi.org//JDS.145