Abstract: China’s coal mining industry is characterized by its high labor intensity, substantial occupational hazards, and elevated accident rates. Although recent stringent national regulations have enhanced the overall safety performance in this industry, major coal mining accidents continue to occur frequently. According to the accident causation theory, coal mining accidents result from the complex interaction of multiple hazards, including people, equipment, the environment, and management. Among these, miners’ unsafe behaviors are the primary cause of accidents. Therefore, exploring how to prevent production accidents caused by unsafe behavior of miners at the source is crucial for enhancing the safety production level in the coal mining industry. In recent years, numerous scholars have conducted extensive research into the mechanisms and prevention strategies of unsafe behaviors. These studies focus on individual-level factors, such as employees’ safety knowledge and psychological capital, as well as organizational-level factors, including safety culture and climate. However, few studies have explored the transmission patterns of individual unsafe behavior in interpersonal contexts and the pathways through which individual unsafe behaviors spread to group unsafe behaviors. In addition, there is often an oversight regarding subject heterogeneity in studies analyzing unsafe behaviors among miners, but failing to differentiate between the types of subjects involved in coal mine safety production. Considering that subject heterogeneity is a critical factor influencing the transmission patterns and effects of miners’ unsafe behavior, further in-depth investigation is required. Based on this, the present study proposes the following two research questions: What roles do different types of subjects in coal mining safety production play in the propagation of unsafe behaviors? Under the condition of subject heterogeneity, how do the unsafe behaviors of various types of coal mining safety production subjects affect others and subsequently lead to the occurrence of group unsafe behaviors?
Current research that relies solely on network analysis and simulation to study propagation lacks effective differentiation and validation of unsafe behaviors among heterogeneous subjects. While existing studies suggest that simple network analysis and simulation can reflect the propagation characteristics of miners’ unsafe behaviors, understanding how individual unsafe behaviors spread and evolve within groups requires further consideration of subject heterogeneity. It necessitates a more in-depth investigation into the propagation of unsafe behaviors among different subject types. Against this backdrop, this study first identifies the unsafe behavior characteristics of various production subjects in 350 coal mine accident reports by the Word2Vec method. A propagation network of miners’ unsafe behavior is then constructed based on association rules and complex network theory. Finally, we identify eight core unsafe behaviors and their associated sets through network centrality analysis. These include: inadequate inspection and supervision of safety management work by government regulators, insufficient daily safety oversight by government regulators, insufficient safety inspections by on-site supervisors, failure to promptly eliminate hidden dangers by on-site supervisors, inadequate safety supervision by on-site supervisors, ineffective safety confirmation by on-site supervisors, and illegal operations and risk-taking behaviors by frontline workers. Moreover, based on the accident path analysis, critical paths for unsafe behavior propagation are identified for six types of accidents: roof, gas, transportation, electromechanical, water hazards, and other types of accidents. The results show that on-site supervisors are the most influential subjects in the propagation of unsafe behaviors among miners. Insufficient safety inspections and failure to promptly eliminate hidden dangers by on-site supervisors are the critical reasons for unsafe group behaviors. The propagation of unsafe behavior most likely causes electromechanical accidents. Inadequate inspection and supervision of safety management work→Inadequate safety education and training→Ineffective on-site safety confirmation→Weak safety awareness among staff→Illegal operations are the critical link for the propagation of unsafe behavior in electromechanical accidents.
This study utilizes text mining and path analysis of 350 coal mining accident reports to identify and integrate the unsafe behaviors of various coal mining safety production subjects. It further reveals the relationships between different types of subjects in the propagation of unsafe behaviors and clarifies their roles in this process. The research implications of this paper can be categorized into three main aspects: Firstly, at the theoretical level, this study reveals the intrinsic mechanisms of the propagation of individual unsafe behaviors to the group under the condition of subject heterogeneity, thereby expanding the related theories and empirical studies on group unsafe behaviors. Secondly, at the methodological level, this study addresses the limitations of traditional network analysis and simulation methods by integrating Word2Vec, association rules, and complex network approaches, which helps to identify the characteristics and propagation mechanisms of unsafe behaviors among different subjects in coal mining safety production. Lastly, at the practical level, this study provides theoretical guidance and a scientific basis for coal mining enterprises to formulate strategies to prevent the propagation of unsafe behaviors. Additionally, this study has certain limitations. It does not cover all subject types involved in coal mining production accidents and focuses mainly on the mechanisms of unsafe behavior propagation across different subjects. Future research should explore the characteristics of populations susceptible to unsafe behavior propagation and develop strategies to prevent the transition of individual unsafe behaviors into collective phenomena.

Key words: unsafe behavior, propagation mechanism, Word2Vec, complex network, association rules

摘要： 群体不安全行为极易导致严重的生产事故,探究个体到群体的不安全行为传播机理对于预防煤矿事故发生尤为重要。为此,本文首先通过Word2Vec方法辨识350起煤矿事故报告中各类生产主体的不安全行为特征;进而结合关联规则和复杂网络理论,构建主体异质性下矿工不安全行为传播网络;最后通过网络中心性分析和事故路径分析明确了8项核心不安全行为及其关联集合、6种事故不安全行为传播关键链路。研究表明:现场监督人员是不安全行为传播过程中最具影响力的主体类型,监督人员安全检查不到位、未及时督促消除隐患是造成群体性不安全行为发生的关键原因。机电事故是不安全行为传播最易触发的事故类型,安全管理工作检查督促不力→安全教育培训不到位→现场安全确认走过场→作业人员安全意识淡薄→现场违章作业是机电事故中不安全行为传播的关键链路。

关键词: 不安全行为, 传播机理, Word2Vec, 复杂网络, 关联规则