Abstract: In order to explore deeply the potential relationship and periodic law in the causal factors of coal mine accidents, the DPM−PHMiner model was constructed to analyze data from coal mine accident investigation reports from 2019 to 2023. A bimodal semantic feature fusion method was employed to identify 30 key factors, which were categorized into five major themes including management and institutional frameworks, equipment and facilities, operational practices, support and protection measures, and safety protection strategies. By using the dynamic pruning−enhanced PHMN+ algorithm, 34 sets of high−efficiency periodic association rules were identified. A multi−dimensional rule evaluation system was established to classify these rules into three priority levels—high, medium, and low—enabling a shift in coal mine safety management from generalized monitoring to targeted risk prevention. The results showed that there were significant differences in the correlation mechanism between the causal factors of different priorities and accidents. High−priority factors, such as illegal operations, exhibited high−frequency, short cycle, and direct trigger correlations with accidents, led rapidly to accident consequences. Medium−priority factors, such as insufficient support strength, were related to accidents in the form of hidden accumulation and gradual development, which was difficult to identify hidden dangers and continued to accumulate until it broke through the critical value under certain conditions. Low−priority factors, including inadequate safety education, demonstrated long−term, systemic, and indirect correlations with accidents. By shaping the background environment where hidden dangers bred, conditions for other causes were created or the effects were amplified, thereby systematically raising the overall risk level.