A Case Study of AutoML for Supervised Crash Severity Prediction

DOI

10.2991/asum.k.210827.026How to use a DOI?

Keywords

Crash severity prediction, Supervised learning, Automated machine learning, Computational intelligence

Abstract

Traffic accidents are one of the leading causes of death around the world. One well-established strategy to deal with this public health issue is the design and deployment of road safety systems, which are in charge of predicting traffic crashes to promote safer roads. Increasing data availability has supported Machine learning (ML) to address the prediction of crashes and their severity. Transportation literature reports various methods for such purposes; however, there is no single method that achieves competitive results in all crash prediction problems. In this context, Automated machined learning (AutoML) arises as a suitable approach to automatically address the model selection problem in areas wherein specialized ML knowledge is not always available or affordable, such as road safety. AutoML has been successfully used in other areas; nevertheless, extensive analysis to determine their strengths and weaknesses has not been done in very diverse learning tasks, such as crash severity forecasting. Thus, this paper aims to examine to what extent AutoML can be competitive against ad hoc methods (Gradient Boosting, Gaussian Naive Bayes, k-Nearest Neighbors, Multilayer Perceptron, Random Forest) on crash severity prediction modeled from a supervised learning perspective. We test 3 state-of-the-art AutoML methods (Auto-Sklearn, TPOT, AutoGluon). Results show that AutoML can be considered a powerful approach to support the model selection problem in crash severity prediction.

Copyright

© 2021, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Juan S. Angarita-Zapata
AU  - Gina Maestre-Gongora
AU  - Jenny Fajardo Calderín
PY  - 2021
DA  - 2021/08/30
TI  - A Case Study of AutoML for Supervised Crash Severity Prediction
BT  - Joint Proceedings of the 19th World Congress of the International Fuzzy Systems Association (IFSA), the 12th Conference of the European Society for Fuzzy Logic and Technology (EUSFLAT), and the 11th International Summer School on Aggregation Operators (AGOP)
PB  - Atlantis Press
SP  - 187
EP  - 194
SN  - 2589-6644
UR  - https://doi.org/10.2991/asum.k.210827.026
DO  - 10.2991/asum.k.210827.026
ID  - Angarita-Zapata2021
ER  -