What is it about?
Using the dataset of Queensland Department of Transport and Main Roads, Australia, we introduce a machine learning based prediction framework to classify road crash severity of vulnerable road users including pedestrian, bicyclist, and motorcyclist. From the Dataset of year 2013 to 2019, 17 distinct road crash parameters were considered as input features to train models, which originate from road user characteristics, weather and environment, vehicle and driver condition, period, road characteristics and regions, traffic, and speed jurisdiction. We predict injury severity of vulnerable road users using machine learning networks: K-Nearest Neighbour (KNN), Support Vector Machine (SVM) and Random Forest (RF).
Featured Image
Photo by Usman Malik on Unsplash
Why is it important?
Our analysis includes a rigorous examination of individual vulnerable road user based prediction and overall predictions using different machine learning approach to identify the best classifier and higher crash severity parameters. Using the best classifier, we measured the crash severity level of different road crash parameters using partial dependency analysis for each individual vulnerable road users and overall vulnerable road users. We plot partial dependency graphs and show the probability of severe crash for different road crash parameters. We find higher severity for motorcyclist.
Perspectives
This paper shows clear picture of Higher Crash Severity in Different Queensland road features and parameters. Our AI based methodology is quite successful identifying better classifier for road crash severity and then measure partial dependency to show higher crash severity in different road crash features based on AI based classifier.
Md Mostafizur Rahman Komol
Queensland University of Technology
The core contribution of this paper is the use of Explainable AI (XAI) to interpret the effects of different parameters and crush severity. Identifying those parameters would help policy-makers understand their vulnerable effects and take possible measures to reduce crush severity for vulnerable road users.
Md Mahmudul Hasan
University of New South Wales
Read the Original
This page is a summary of: Crash severity analysis of vulnerable road users using machine learning, PLoS ONE, August 2021, PLOS,
DOI: 10.1371/journal.pone.0255828.
You can read the full text:
Contributors
The following have contributed to this page
