ESTIMATION OF COST IMPACT OF RISK OF INJURY FROM CONSTRUCTION ACCIDENTS USING LONG SHORT-TERM MEMORY NEURAL NETWORK ARCHITECTURE, BASED ON PARTS OF THE BODY AFFECTED

Recurrent neural networks (RNNs) are increasingly being used to analyse occupational accidents related to construction projects to inform project stakeholders about the effects of accidents on project costs. Thus, this study used Long, Short-Term Neural Network Architecture (LSTM) to develop a model for predicting the cost of construction-related occupational accidents based on the body parts affected. Pycharm Community 2020.1.2 (using Python 3.9.6) was used for the experiment and analysis, in this experimental study. The findings effectively showed that the suggested LSTM model's prediction performance, with a training and testing ratio of 90:10, was determined to have a mean square error (MSE) of 0.003 and a root mean square error (RMSE) of 0.056, respectively. Thus, these demonstrate that the LSTM5 model is a viable substitute that can help government agencies, contractors, clients, project managers, and health and safety managers estimate the cost impact of risk of injury resulting from construction-related accidents for any building construction project. The model would help in accurately estimating the risk impact and mitigation processes.