A HYBRID ROAD SAFETY EVALUATION-PRIORITIZATION APPROACH (RSE-PA) FOR INTERCITY ROAD NETWORK IN TURKEY

Date

2024-4-25

Author

Topçuoğlu, Baran

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

104
views

62
downloads

Cite This

Based on total number of accidents, the General Directorate of Highways (GDH) is obliged to detect hazardous locations and make interventions, which require further engineering studies (i.e. speed studies, conflict studies, etc.). A road safety evaluation (RSE) program can be used to address potentially problematic sub-problems such as speed-related accidents, and prioritize their locations. This study proposes a hybrid RSE approach which combines clustering and cause-and-result approaches, as the first step of future risk assessment model for GDH network. The clustering analysis part uses both DBSCAN and Nearest Neighborhood Hierarchical (NNH) techniques using geocoded accident data. Digitization of the road network into equally sized slices further allowed i) assignment of both geometry data (i.e. median width, availability of lighting etc.) and traffic characteristics (i.e. volume, speed, heavy vehicle number and ratios, etc.), which are later used to derive Decision Tree (DT) models estimating factors explaining traffic accidents occurences on highway slices. Using speed-related accidents between 2018-2022 and relevant data from the GDH Mersin Region and selected corridors of Ankara intercity highway corridors, showed that NNH and DBSCAN detect clusters which mostly overlaps (partially or fully). DT analysis of 250 m-long slices also produced statistically significant models predicting probability of occurrence of accident with upto 70.5% reliability. Combining the clustering and DT results in a total scoring approach for each road slice, leads to selection of many road slices overlapping with major NNH and DBSCAN clusters extended by nearby slices carrying similar traffic and geometric features detected risky by DT results.

Subject Keywords

Traffic Accidents, Spatial Clustering, Road Safety Evaluation, DBSCAN, GIS

URI

https://hdl.handle.net/11511/109800

Collections

Graduate School of Natural and Applied Sciences, Thesis