Year: 2025
Type: conference-paper
Abstract: Global estimates indicate that the construction industry has one of the highest rates of occupational accidents. Given this scenario, the sector has adopted emergent technologies to improve safety conditions and decision-making. Thus, machine learning (ML) emerges as a promising tool to streamline data analysis besides being used to predict events based on datasets. In construction, the use of the ML model to predict site accidents remains emerging without studies using historical data from emerging South and Latin American countries. Therefore, this paper aims to develop a prediction model using machine learning for fatal accidents using historical data from construction. The dataset contains 2,305 accidents, including fatal and non-fatal, from 2018 to 2023, distributed in 17 binary and categorical variables. After the pre-processing, seven predictive models were generated with different classifiers to determine which models fit the dataset better. Among the classifiers, the gradient-boosting model exhibited the best performance, achieving an accuracy of 0.885, 0.88 precision, 0.833 recall, and 0.881 F1-Score. As a contribution, this study brings insights regarding the use of predictive models in construction safety management, calling attention to the attributes that have the most influence on a fatal accident. In addition, the model can assist managers in identifying scenarios that are more prone to accidents and propose effective preventive measures.
Source: CIB Conferences
Authors: Filipe dos Santos Freitas, Mirian Caroline Farias Santos, Roseneia Rodrigues Santos Melo, Paulo Henrique Ferreira, Dayana Bastos Costa
Institution: Universidade Federal da Bahia
Language: English
Subfield: Radiological and Ultrasound Technology
Field: Health Professions
Domain: Health Sciences
Open Access status: bronze
OpenAlex ID: W4411471415