Application of Bayesian network model in explaining occupational accidents in a steel industry

In the occupational accident analysis, identification of the interrelationships of the factors behind the accidents is very important. To explore the relationships or the impacts of the causal factors on the accidents and to predict the incident outcomes i.e., injury, near miss, and property damage cases, Bayesian Network (BN) model is used in this paper. The proposed model is validated using the data retrieved from an integrated steel manufacturing industry in India using sensitivity analysis. BN performs well in terms of prediction with 88.28% accuracy using 10-fold cross validation. In addition, some important key findings are obtained from the analysis like the factors slip-trip-falls, crane dashing, and the months February and July are found to be the sensitive factors towards incident outcomes in the plant. The proposed model, therefore, has a good potential for explaining accident prediction and causation in manufacturing industry and can be applied in different domains also.