A Comprehensive Study of Road Traffic Accidents: Hotspot Analysis and Severity Prediction Using Machine Learning

Abstract

This study analyses road traffic accident data recorded over a period of time to gain insights to the underlying pain points in the infrastructure and policies. Such insight allows us to focus our efforts in the right direction to make the lives of people safer. The data includes various geographical and meteorological factors affecting the severity of these accidents. We use Kernel density estimation (KDE) plots to analyse hotspots of accident-prone areas weighed against severity over years to understand the evolution of these dangerous zones. Furthermore, we use machine learning algorithms to predict the accident severity given certain parameters and to understand the factors that have a major influence on the severity of the accident. We have studied a publicly available dataset of road traffic accidents in the UK as a proof of concept of the pipeline to understand the underlying patterns of accidents occurring in a region of interest.