Date

2022-8-15

Author

Özdemir Aydın, Merve

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

232
views

173
downloads

Cite This

Underground coal mining is considered among the most dangerous sectors in the world due to the accidents. Thus, this study aims to build an accident severity prediction model for underground coal mines by using decision tree, support vector machine, and neural network algorithms. Defining the severity of accidents will provide an effective way of preventing risks that will cause serious accidents. This study also aims to fill the gap in the literature related to designing accident severity prediction models for underground coal mining for safety management. In the study, 8406 underground accident data covering two years period of time, and eleven variables (dimensions), which are shift, day of the accident, job, education, type of accident, reason of the accident, location of the accident, severity of the accident, age, seniority, affected body part, collected by the Turkish Hard Coal Enterprise of Amasra, Armutçuk, Karadon, Kozlu, and Üzülmez district were used to build an accident severity prediction model. Before applying the machine learning algorithms, principal component analysis was applied to reduce the dimensions and express the data with fewer variables that are meaningful and easier to explain. Principal component analysis provided that 81.82% (cumulative variance percent) of the data could be interpreted with the seven components. By using these seven variables, accident severity prediction models were built applying decision tree, support vector machine, and neural network algorithms. The decision tree model has the accuracy 78.5%, support vector machine model has the accuracy 79.2%, and neural network model has the accuracy 78.5%. As a result, it was decided that the accident severity estimation model that gives the most accurate prediction results is the support vector machines for this data set. Based on trained prediction model results, the dominant correct classification accident severity type is slightly injured.

Subject Keywords

Underground coal mine, Principal component analysis, Decision tree, Support vector machine, Neural network

URI

https://hdl.handle.net/11511/98630

Collections

Graduate School of Natural and Applied Sciences, Thesis