Article

Identifying Homeless Shelter Archetypes via K-Means Clustering to Inform State Responsibility 

---

Abstract

Public policies designed to address homelessness often rely on broad, system-wide metrics that mask significant operational heterogeneity, leading to inefficient "one-size-fits-all" strategies. This paper argues that the availability of granular data and accessible machine learning techniques creates a new standard for state responsibility under international human rights law. We employ an unsupervised machine learning approach, specifically K-Means clustering, to analyze a dataset of daily operational metrics from a network of homeless shelters. The analysis is based on four key features: total capacity, occupancy rate, average age, and the percentage of male occupants. The clustering algorithm successfully identified four distinct and interpretable shelter archetypes, revealing a hidden typology within the system. The most critical finding is the emergence of a "Strained Mid-Sized Shelter" archetype, characterized by moderate capacity and the highest average occupancy rate, providing empirical evidence of a recurring state of systemic stress. The existence of these data-defined archetypes transforms the abstract risk of shelter failure into a concrete and foreseeable event. We conclude that this data-driven foreseeability elevates the state's duty to act under the Right to Adequate Housing (UDHR, Article 25). The failure to use available analytical methods to identify and respond to predictable patterns of strain can be construed as a breach of this duty. This study provides a novel framework for linking data science to legal accountability, advocating for the adoption of evidence-based, targeted policies that reflect the nuanced realities of social service provision. This approach offers a new paradigm for holding states accountable for protecting the rights of vulnerable populations in the digital age.

Keywords: Clustering, Foreseeability, Homelessness, Human Rights, Public Policy

How to Cite:

Budiyanto, H. & Angelia, C. R., (2025) “Identifying Homeless Shelter Archetypes via K-Means Clustering to Inform State Responsibility ”, Journal of Cyber Law 1(3), 228-245. doi: https://doi.org/10.63913/jcl.v1i3.46