Development of prediction model of construction workers accident occurrence through machine learning

Abstract

In the Republic of Korea, the industrial accident rate and the number of casualties in the construction industry have continued to increase for six years from 2011. The average industrial accident rate is 26.44% and the average number of casualties is 24,183. To prevent accidents, the Ministry of Employment and Labor (MOEL) presents various analysis data through the annual industrial accident report, but this has not been effective in reducing accidents as a result. In this paper, using the Logistic regression model that is one of the Machine learning method, this study develops a construction accident prediction model by training 80% of the data of accident casualties (25,114 persons) and accident deaths (499 persons) by 2016 and tests the predicted model with 20% unused data. And then, this study presents a construction site safety management process using the predicted model. The model and process developed in this paper are expected to contribute to the safety management of the construction site as a tool to prevent fatal accidents of construction workers.