Evaluating the performance of GUIDE, MOB, and CART in traffic accident prediction: A case study from Ankara
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Traffic accidents on state roads represent a major concern, resulting in injuries, fatalities, and significant economic losses. Identifying the primary factors that contribute to these accidents is vital for developing effective prevention measures. This study examines the frequency of traffic accidents on state roads in Ankara by employing Poisson regression and regression tree models, focusing on the comparison of three regression tree algorithms. A detailed simulation study was carried out to assess the performance of the tree algorithms in terms of variable selection bias, Type I error rates, and statistical power. Among them, GUIDE demonstrated the most balanced performance, with unbiased variable selection and strong power, effectively controlling Type I errors. In contrast, the CART algorithm outperformed others in scenarios involving overdispersion. Both CART and GUIDE exhibited similar estimation errors in real-world applications, highlighting their robustness and reliability. Additionally, the study observed that the MOB algorithm tended to favor numerical variables over categorical ones when performing splits. Although it maintained adequate control over Type I errors, MOB showed relatively lower power, reflecting its limitations in data partitioning. In conclusion, the findings offer valuable insights into selecting appropriate regression tree algorithms for traffic accident analysis, providing guidance for enhancing road safety.
Fields of Study & Subjects
Traffic and Road Safety
Statistical Methods and Inference
Statistical Methods and Applications
Publication details
- Journal / source
- Communications in Statistics Case Studies Data Analysis and Applications
- Pages
- 1–21
- Publication date
- 26 March 2026
- Language
- EN