As a high-risk occupation, coal mining has many accidents, primarily due to the unsafe behavior of coal miners. Based on the research of analysis of unsafe behavior and pan-scenario data of miners, a theoretical framework for the analysis of unsafe behavior characteristics was proposed in this paper. The collected data were divided into realistic scenes and abstract scenes according to different manifestations; the pan-scene data were described from the eight dimensions of time, behavioral trace, location, behavioral property, behavioral individual, degree, unsafe action, and specialty using a quantitative method for the structure conversion; and the rules were discovered through cluster analysis and association analysis. A total of 225 coal mine gas explosion accidents were used for analysis, and the pan-scene data description and structure conversion of unsafe behavior that caused these accidents were realized. In a certain cluster, the distribution rules of dimensions and the interaction between different dimensions of unsafe behavior were explored after analysis. The results show that the proposed eight dimensions can fully explain the basic characteristics and attributes of the unsafe behavior of coal miners. The structure conversion can reduce the workload of managers and effectively improve the safety data processing capabilities, and the result of data analysis can provide data support and a management basis for safety management. A new method and thought for the data analysis of miners’ unsafe behavior is provided.