Feature Policy: A hybrid approach to minimize the severity of road crashes through interventions in predictor variables

Abstract

Introduction: Road crashes are a major public health concern, leading to numerous fatalities and substantial economic costs. Reducing these incidents is a key priority outlined in the United Nations Sustainable Development Goals. Existing research often identifies contributing factors, analyzes relevant variables, and proposes preventive strategies for crashes using logistic regression (LR), but lacks integrated policy optimization. Method: This study presents a Feature Policy Optimization (FPO) framework combining LR and a genetic algorithm (GA) to optimize speed limits and lighting conditions. The LR model estimates crash risk based on these factors, while the GA generates Pareto-optimal policies under constraints of budget and a minimum average speed of 50 km/h. Policies are tailored by road type, time, and location. Results: The FPO approach reduced serious crashes by 40.58% compared to baseline, and identified critical influences (such as nighttime lighting on rural roads), and recommended adaptive speed limits that balance safety and traffic flow. Conclusion: Integrating LR with GA effectively identifies and modifies key accident factors, enabling targeted, cost-effective safety interventions. Practical applications: The FPO model offers policymakers actionable, context-specific strategies for improving road safety. Its scalable framework supports implementation across diverse transportation settings, contributing to sustainable crash reduction.