Predictive modeling of workplace accident outcomes utilizing XGBoost and Tree-Structured Parzen Estimator
In this study, workplace accident outcomes were predicted using Tree-Structured Parzen Estimator optimized XGBoost algorithm. Outcomes were divided into serious and non-serious accidents based on the absence from work. Despite its limitations, the TPE-optimized model could predict serious accidents with an accuracy of 73% and non-serious accidents with an accuracy of 77%.
In this research, we used the TPE method to optimize the XGBoost model to predict workplace accident outcomes based on the accident notices delivered to the insurance companies in Finland. Accident outcomes were divided into serious and non-serious accidents based on the absence from work resulting from the accident. Cases where the absence from work was more than 30 days, were considered serious.
Wounds and superficial injuries and bone fractures were found to be the most important features predicting the workplace accident outcome with wounds and superficial injuries, implying in most cases a non-serious accident and bone fractures serious accident. Overall injury variable was found to be most important ESAW-variable together with body
The model could predict serious accidents with an accuracy of 73% and non-serious accidents with an accuracy of 77%.