Identification and analysis of offenders causing hit and run accidents using classification algorithms

Area of Science:

• Traffic Safety
• Machine Learning Applications
• Accident Analysis

Background:

• Hit-and-run crashes pose a significant global challenge due to difficulties in identifying offending vehicles.
• Lack of data on impacting vehicles hinders understanding crash dynamics and developing effective prevention strategies.

Purpose of the Study:

• To identify the types of vehicles involved in hit-and-run (hit-and-run) crashes.
• To develop a robust and scalable framework for identifying impacting vehicles in such incidents.

Main Methods:

• Analysis of fatal road crashes in New Delhi from 2006 to 2016.
• Application and comparison of eleven machine learning classification algorithms, including Support Vector Machine with linear kernel (SVM-linear-kernel).

Main Results:

• Approximately 40% of fatal crashes were identified as hit-and-run incidents with unknown impacting vehicles.
• The SVM-linear-kernel algorithm demonstrated the best performance in identifying vehicle types.
• Cars, buses, and heavy vehicles were found to be the primary vehicles involved in hit-and-run crashes.
• Buses were the main cause of hit-and-run crashes from 2006-2009, with a subsequent increase in car involvement.

Conclusions:

• The developed machine learning framework is effective and scalable for identifying vehicles in hit-and-run crashes across different cities.
• Findings provide valuable insights for traffic engineers to develop targeted policies and enhance road user safety.