Prediction of number of insured having work accident in Turkish construction industry: a neural network fitting approach

Abstract

Accurate forecasting of work accidents is of paramount importance in promoting workplace safety and improving risk management strategies. This study proposes a novel approach based on a neural network fitted with the Levenberg-Marquardt algorithm to predict future accident numbers in 22 different occupational groups within the Turkish construction industry. By utilizing historical official data spanning the years 2014 to 2022, the aim is to provide insights into the potential accident rates that may arise in different job categories. The constructed neural network model consists of two hidden layers. Leveraging the powerful capabilities of the Levenberg-Marquardt algorithm, the network is trained to effectively capture the complex dynamics underlying work accidents in the construction industry. The findings demonstrate the effectiveness of the proposed approach in forecasting future accident numbers with a high degree of precision. The neural network model successfully leverages the temporal trends and underlying factors present in the historical data. By employing an advanced neural network framework and the Levenberg-Marquardt algorithm, this study offers a robust methodology for predicting work accident rates across diverse job categories. The results obtained from this study can guide the development of targeted preventive measures, tailored training programs, and efficient resource allocation strategies.