Investigation of factors affecting occupational accidents in coal mining with data mining

The mining sector is one of the most high-risk work environments by nature and has the highest frequency of occupational accidents worldwide. Occupational accidents not only lead to workforce and financial losses but also cause serious injuries and fatalities. In this thesis study, data from occupational accident notification forms in the mining sector obtained from the Social Security Institution (SGK) for the years 2013–2020 were analyzed to determine the parameters affecting the severity of injury-related accidents in underground coal mining. To achieve this, the performances of models using data mining methods—artificial neural networks, decision trees, and support vector machines—were compared with the help of the SPSS Modeler program. Since the artificial neural networks model provided stronger predictions, the effects of subcategories of variables on accident severity were examined in more detail. In the analyses, it was concluded that the region, equipment and general activity parameters were the most important variables according to the first model. A second model consisting of a total of 46 variables was established by including all subcategories of the parameters used in the first model. When the importance levels of the subcategories of the parameters in this model were examined, it was determined that the Black Sea region had the highest importance, and excavation, construction, repair, demolition works and works on buildings, structures and surfaces had high importance. Subcategories such as movement and age groups had limited effects on accident severity. In this context, a basin-oriented mining management approach should be adopted—one that considers regional risk profiles to ensure a sustainable and safe working environment.