From planning to perception: A study of urban security in new towns using machine learning, spatial clustering, and lived experiences

Urban security in new towns is a multifaceted challenge demanding comprehensive Crime Prevention Through Environmental Design strategies. This study aimed to analyze the perception of security and its determinants across three Iranian new towns: Pardis, Andisheh, and Parand. A novel mixed-methods approach was employed, integrating Hierarchical Clustering to delineate security profiles, the Random Forest model to identify the strongest predictors, and Thematic Analysis to capture residents' lived experiences. The results reveal significant variation in security dynamics and their drivers: Pardis exhibits the highest security, where Social Cohesion is identified as the dominant predictive factor. Conversely, Parand reports the lowest perceptions, driven by a systemic failure in governance and infrastructure, resulting in an over-reliance on basic Territorial Reinforcement cues. Furthermore, Andisheh demonstrates an 'Efficacy Conflict,' where positive design features are undermined by institutional weakness and social fracturing, preventing them from translating into effective security. The study underscores the necessity of moving beyond rigid CPTED frameworks to adopt a hybrid, context-specific strategy that holistically integrates environmental design, social capital, economic stability, and effective governance to build safer and more resilient urban communities. • Uses Random Forest, clustering, and thematic analysis to assess urban security. • Develops a hybrid CPTED model for towns with weak cohesion and governance. • Finds Pardis secure via cohesion, Andisheh conflicted, and Parand institutionally weak. • Shows social capital and governance are as vital as physical design for security. • Proposes a context-based model linking design, community, and data-driven policy.