Title: Traffic accident analysis and development of accident prediction model for four-lane divided national highway

Authors: Vinod Kumar; Sanjeev Kumar Suman

Addresses: Department of Civil Engineering, NIT Patna, Patna, 800005, Bihar, India ' Department of Civil Engineering, NIT Patna, Patna, 800005, Bihar, India

Abstract: This study analyses traffic incidents on divided four-lane national highways to identify trends and develop a novel accident prediction model. By applying the weighted severity index (WSI) method, the research highlights accident-prone areas (dark spots) based on both frequency and severity, guiding targeted safety interventions. The research employs a cutting-edge approach by integrating geometric and traffic parameters using the CGAN-EB methodology, a non-parametric Empirical Bayes method leveraging conditional generative adversarial networks (CGANs) that provide a dynamic and data-driven prediction model. The model, incorporating safety coefficients from the analytic hierarchy process (AHP), contributes to improved highway design and safety measures. MATLAB can be a powerful tool for conducting traffic accident analysis and developing accident prediction models for divided four-lane National Highways, achieving a high accuracy of 98% in a predictive model for traffic accident analysis.

Keywords: traffic; accident black spots; WSI; weighted severity index; Empirical Bayes; CGANs; conditional generative adversarial networks; AHP; analytic hierarchy process.

DOI: 10.1504/IJHVS.2025.150208

International Journal of Heavy Vehicle Systems, 2025 Vol.32 No.6, pp.798 - 822

Received: 26 Nov 2024
Accepted: 06 Jan 2025
Published online: 03 Dec 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article