"Comparative Analysis of Route Optimization Algorithms for Urban Safety: Implications for Public Policy in Latin American Cities"

Abstract

"This study presents a comprehensive evaluation of safe route selection algorithms for urban environments, comparing Deep Graph Networks (DGN) combined with Deep Reinforcement Learning (DRL) against five traditional algorithms using Los Angeles crime data (2020-2024, 925,000 records). While DGN+DRL shows theoretical promise for handling complex spatial relationships, empirical results reveal that simpler algorithms, particularly Monte Carlo (12.63% crime index), outperform the complex deep learning approach (16.25 % crime index, 29 nodes vs. 16 nodes). These findings have critical implications for public policy: (1) open crime data enables evidence-based urban planning, (2) algorithm selection must balance complexity with performance, and (3) traditional methods may be more practical for immediate deployment in resource-constrained Latin American cities like Lima (crime index 70.9%). The study provides actionable insights for policymakers implementing smart city safety initiatives, demonstrating that algorithm effectiveness depends more on problem formulation than model complexity. Results suggest that traditional approaches (Dijkstra, A*, Monte Carlo) with processing times under 2 seconds provide 85% of desired safety improvements at one-tenth the computational cost, making them viable solutions for real-time mobile applications in cities with limited technological infrastructure © 2025 IEEE."