Analyzing Spatial Heterogeneity in Road Traffic Accidents Using Geographically Weighted Regression: A Localized Approach for Targeted Safety Interventions

Abstract

Road traffic accidents remain a critical global public health issue, necessitating advanced analytical approaches that account for spatial heterogeneity in accident data. Traditional methods, such as Ordinary Least Squares (OLS) regression, often overlook localized variations, resulting in suboptimal policy interventions. This study applies Geographically Weighted Regression (GWR) to analyze accident patterns in Jega, Nigeria, capturing spatially varying relationships between accident occurrence and factors such as road design, traffic volume, and land use. The GWR model demonstrates strong explanatory power, with a global R² of 0.72 and an AICc of 420.35, outperforming traditional regression techniques. Cross-validation results (RMSE = 3.45, MAE = 2.12, R² = 0.65) confirm the model’s predictive robustness and generalizability. Spatial variability in local R² values (0.20–0.95) underscores the importance of localized analysis, with clearly identified high-risk zones enabling targeted interventions. The study highlights GWR’s effectiveness in addressing spatial non-stationarity and offers actionable, geographically specific insights for policymakers and planners to enhance road safety. Challenges including multicollinearity and computational demands are acknowledged, and future research directions such as hybrid modeling and real-time data integration are proposed