Text Mining Analysis of U.S. Department of Labor’s MSHA Fatal Accident Reports for Coal Mining

"Coal mining is an age-old method of power production in the world. It is also widely known for the associated risks during mining. Although the number of mishaps has dropped dramatically over the years, the number of fatal accidents is still a concern, and any aid in reducing these figures would be important to industry workers. Therefore, data from accident reports during the past seven years were collected and analyzed. Qualitative data from reports and articles provide a deeper understanding of the incidents, thus affording the possibility of identifying patterns and relationships between the accidents. For this study, a total of 119 fatality accidents occurring in U.S. coal mining operations were analyzed using advanced text mining techniques. The fatality data were obtained from U.S. Mine Safety and Health Admininstration fatalgram reports. This analysis is primarily exploratory in nature and uses term frequencyinverse document frequency (TF-IDF) methodologies along with correlation network plots. An interesting word relation pattern was obtained that identified “positioning” as the common cause of 12.6 percent of accidents, and found that 29.4 percent of accidents can be categorized as vehicle-related. IntroductionIn the mining industry, accidents are situations that employees and employers wish to avoid. They are devastating experiences to the victims and their coworkers alike. Even worse, accidents that result in fatalities create trauma and undesirable repercussions. Several measures have been taken throughout the industry to prevent such mishaps, and the U.S. Mine Safety and Health Administration (MSHA) provides continuous monitoring for miners’ safety by guiding organizations through best practices in every job role. As a result, according to MSHA, there has been a steady decline in the average number of deaths due to accidents over time. Nonetheless, any opportunity to avert one more such casualty is of paramount importance, and that is the goal of our analysis. Our objective is to identify opportunities resulting from previously unexplored directions in order to provide additional insights into potential safety recommendations. Until now, most of the studies in this area had been conducted using structured and tabulated data with a constant number of categorical and numerical variables. Unstructured text data mining is an emerging area that is still insufficiently researched and has a feasible scope for important results, as a vast majority of human communication and cognizance is comprised of words. The aim of undertaking this project by capitalizing on unstructured data is to analyze the text reports to discover potential relationships that explain the causes of accidents in new light in order to mitigate the occurrence of fatalities. The analysis in this project is based on textual data obtained from the MSHA website (2017). Individual accident reports are organized according to the date and year of the fatality. Each report includes details with images of the incident. For convenience, a simplified summary of the report is also available. Data collection consisted of gathering formal documentation of fatal accidents in the coal mining industry during the past seven years. These reports were extracted from the website in text format."