Macro-level accident modeling in Novi Sad: A spatial regression approach

Area of Science:

• Urban Planning
• Transportation Engineering
• Spatial Analysis

Background:

• Traffic accident factors vary significantly by geographic area, creating a gap between research findings and practical application.
• Understanding these localized factors is crucial for developing effective road safety interventions.
• A three-year dataset from Novi Sad was used to investigate accident determinants.

Purpose of the Study:

• To identify macroscopic factors influencing traffic accidents within defined traffic analysis zones.
• To compare the performance of different spatial predictive models in accident analysis.
• To provide data-driven insights for improving road safety in Novi Sad.

Main Methods:

• Macroscopic analysis of traffic accidents over a three-year period.
• Application of spatial predictive models: classical predictive space model, spatial lag model, and spatial error model.
• Examination of total accidents, motorized, and non-motorized mode accidents.

Main Results:

• The spatial lag model best predicted total and motorized accidents, while the spatial error model excelled for non-motorized accidents.
• Increased Daily Vehicle-Kilometers Traveled (DVKT), parking spaces, 5-legged, and signalized intersections were linked to higher accident rates.
• Demographic, traffic, road, and environmental characteristics showed varied impacts on different accident types.

Conclusions:

• Spatial models effectively identify accident contributing factors, with model performance varying by accident type.
• Key infrastructure elements like intersections and parking, alongside traffic volume, are significant accident predictors.
• Findings support targeted interventions in urban planning and traffic engineering for enhanced road safety.