EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

4200. Prioritization of traffic crash hotspots for intervention

Invited abstract in session WC-55: Big data analysis and AI in transportation, stream Transportation.

Wednesday, 12:30-14:00
Room: S02 (building: 101)

Authors (first author is the speaker)

1. Yehezkel Resheff
Business School, Hebrew University
2. Mali Sher
Trafic Police, Israel Police
3. Nicole Adler
School of Business Administration, Hebrew University of Jerusalem

Abstract

Traffic crashes remain a major cause of preventable loss of life and injury. Many policy and intervention efforts aim to reduce the number and severity of collisions. Of these, the most direct is the identification and correction of hotspots of severe crashes. We present data from the 20 largest cities in Israel, spanning over a decade, showing that locations of clusters of light and property damage crashes have a risk of a severe crash in the following year that is comparable to that of clusters of severe crashes. This has immediate policy implications, as we should rethink the intervention strategy that currently predominantly considers severe hotspots. Next, we apply machine learning to model the probability of having a future severe event at a cluster, given the known characteristics of the location and the previous accidents that occurred. The quality of the model is shown to differ between types of clusters, such as severe vs. minor, intersection vs. non-intersection, etc. Finally, the problem facing a decision maker is how to use a given budget and prioritize clusters for intervention, based on the risk scores provided by the model and their type-dependant uncertainties, as well as the type-dependant cost of intervention. The overall aim of the policy is assumed to be maximization of the total reduction of expected future severe crashes, but extensions that consider fair spread of resources are also considered.

Keywords

Status: accepted

Back to the list of papers