An approach to estimate occupational accidents using least-squares support vector machines

Abstract

Least-squares support vector machines represent an emerging technique that has been adopted to estimate accidents. In this study, occupational accident estimation models were developed using the least-squares support vector machine method for the Republic of Turkey. In addition, linear regression analysis, nonlinear regression analysis, and artificial neural network models were considered. During the development phase of the models, statistical data from 1970 to 2012 were used to consider parameters such as insured workers, workplaces, occupational accidents, deaths, and permanent incapacities. Using these models, the numbers of accidents, deaths, and permanent incapacities resulting from occupational accidents were estimated for three different scenarios in the Republic of Turkey through the end of 2025. The performances of the developed models were evaluated considering the mean absolute percent errors and the mean absolute errors. In addition, we compared the least-squares support vector machine, linear regression analysis, nonlinear regression analysis, and artificial neural network methods in terms of their estimation performances. Our simulation results demonstrate that the proposed least-squares support vector machine model outperforms other techniques in terms of accuracy and has a rapid convergence capability when estimating occupational accidents.