Investigation of ocupational accidents in Amasra hard coal enterprise

Nowadays, as in every branch of industry, a large amount of data can be collected in mining, both in productivity and occupational safety. It is increasingly essential to transform this data into useful information for enterprises. Data mining is very useful in processing and extracting useful information from the processed data. This thesis aims to analyze the data of occupational accidents with injuries between 2010 and 2021 in Amasra Hard Coal Enterprise of Turkish Hard Coal Corporation by data mining. For this purpose, the injured accident data for the relevant years were taken from the enterprise database and organized in a way suitable for the study. ZeroR, J48 Decision Tree, k Nearest Neighborhood, Artificial Neural Networks, Naive Bayes, Support Vector Machine, and Random Forest data mining algorithms were used in the evaluation phase. These algorithms were applied with the WEKA program, an open-source application. According to these results, it was tried to determine the algorithms that best classify and predict accident data for the enterprise. According to different test methods, k Nearest Neighborhood and Support Vector Machine algorithms succeeded in classification and prediction. One of the critical phases of the study is the determination of the attributes and subclasses that are effective in the origin of accidents by association rules mining. Thus, more detailed information was obtained about the locations and root causes of the accidents that occurred in the enterprise. As a result of Apriori and Predictive Apriori applications, it was revealed that the root causes of occupational accidents according to the accident locations are the experience of the workers, the working hours in the shift, and the worker position. In addition, shift hours, accident cause, monthly production, and monthly wage variables were also influential. Given these results, recommendations for the enterprise are listed.