Labor Accidents Classification Using Machine Learning

Enádio Barbosa, Yandre Costa, Juliano Foleis, Diego Bertolini

2023

Abstract

The application of artificial intelligence is increasingly growing in all public and private industry fields. In this work, we propose applying machine learning techniques to perform work accident classification according to Brazilian laws. The type of accident is part of the communication of occupational accidents (CAT) database held by the National Institute of Social Security. In Brazil, that communication can come from different sources. Because of this, some of them lack the type of work accident. This information is crucial to allow labor authorities to understand better the circumstances surrounding the accidents and to help plan and create more specific strategies to prevent them. In this sense, we perform data cleaning, and we use feature engineering techniques to treat data from CAT database. Following, we use machine learning algorithms aiming to perform the classification according to the type of accident. For this, we investigate a strategy to identify the type of labor accident when this information is missing using algorithms based on trees or gradient boosting trees. Preliminary results showed promising performance, where the algorithms achieved the following weighted average F1-score for labor accident types classification: XGboost 0.94, CAtboost 0.94, Lightgbm 0.94, and Random Forest 0.91. As far as we know, work accident type classification using machine learning, considering Brazilian labor legislation and a huge governmental dataset is addressed for the first time in this work.

Paper Citation

in Harvard Style

Barbosa E., Costa Y., Foleis J. and Bertolini D. (2023). Labor Accidents Classification Using Machine Learning. In Proceedings of the 25th International Conference on Enterprise Information Systems - Volume 1: ICEIS, ISBN 978-989-758-648-4, SciTePress, pages 509-516. DOI: 10.5220/0011856500003467

in Bibtex Style

@conference{iceis23,
author={Enádio Barbosa and Yandre Costa and Juliano Foleis and Diego Bertolini},
title={Labor Accidents Classification Using Machine Learning},
booktitle={Proceedings of the 25th International Conference on Enterprise Information Systems - Volume 1: ICEIS,},
year={2023},
pages={509-516},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0011856500003467},
isbn={978-989-758-648-4},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 25th International Conference on Enterprise Information Systems - Volume 1: ICEIS,
TI - Labor Accidents Classification Using Machine Learning
SN - 978-989-758-648-4
AU - Barbosa E.
AU - Costa Y.
AU - Foleis J.
AU - Bertolini D.
PY - 2023
SP - 509
EP - 516
DO - 10.5220/0011856500003467
PB - SciTePress