Identifying coal mine safety production risk factors by employing text mining and Bayesian network techniques

Coal industry is a typical high-risk industry with frequent accidents. In an effort to ensure workers’ safety and health, and reduce the probability of productivity decrease, it is essential to identify the contributing factors of coal mine safety production risks through certain technical means. Accident cases, as a concentrated display of accident hazard source, are of great value in extracting key risk factors that may induce coal mine disasters. Therefore, this study creatively proposed an effective method combining text mining, association rule mining and Bayesian network to deeply mine and use the massive coal mine safety accident case text data, so as to achieve effective identification of coal mine safety risk factors and explore the mechanism of interaction between risk factors and their importance. The research main included three steps. First, due to the high uncertainty and difference in the way of expression of the coal mine accident report texts, the conventional text mining process cannot effectively identify the risk factors, resulting in the incompleteness and deviation of the risk factors list. This study improved the text mining process, through Chinese word segmentation, keyword extraction, related word mining, semantic analysis, etc. to mine the collected 726 reports, and identify 78 safety risk factors. Then, the Apriori algorithm was used to obtain the extremely frequent itemset of risk factors and 362 strong association rules, and constructed the Bayesian network model on this basis. Finally, six main risk factors of coal mine safety production and their associated-factors were clarified through sensitivity and critical path analysis. The study shows that compared with the risks caused by the environment and equipment, the lack of management, education, and supervision are the root cause of coal mine accidents. This research provides a new way of thinking for effectively extracting using information from unstructured and non-standardized texts, as well as a new perspective for data-driven safety risk factors identification and complex interaction mechanisms research, having a great significance for coal mine safety risk pre-control management.