**5. Conclusions**

The research of targeted interventions in other industries and different fields were combined with behavioral safety managemen<sup>t</sup> and applied to unsafe behavior interventions. By dividing the data boundary, using safety managemen<sup>t</sup> data in coal mine production and the method of single-dimensional statistical analysis and multi-dimensional interactive analysis, the targeted intervention process of miners' unsafe behavior was constructed and deemed suitable for coal mines. In the study of miners' unsafe behavior, this process realizes innovation from accident cases to data analysis and to behavior intervention. The method and process of the data analysis of behavior promote further development in the field of behavior intervention, and lay the theoretical foundation and method reference for the realization of accurate identification, accurate intervention, and accurate managemen<sup>t</sup> of unsafe behavior.

The coal miners' unsafe behavior was taken as the target intervention object. Through the analysis of the data of the coal mine roof accident, the characteristics of data distribution and the interaction rules in the accident case were excavated, and the target intervention nodes of the unsafe behavior were then located. The targeted intervention nodes can help to find unsafe behaviors that can easily lead to roof accidents, and provide a direction for enterprises to reduce the incidence of unsafe behaviors more quickly. According to the different proportion of intervention provided by intervention nodes, enterprises can optimize the allocation of managemen<sup>t</sup> resources and improve managemen<sup>t</sup> efficiency. Multiple intervention strategies provided by multiple intervention nodes have an effectively practical significance in controlling roof accidents in coal mines.

The evaluation model of targeted intervention nodes was established by using the structural equation method (SEM), based on which the evaluation process of unsafe behavior-targeted intervention nodes was formed. According to the evaluation results, the effectiveness of each target intervention node was determined, and the scientific and practical nature of the proposed method was further verified.

Based on the roof accident data, we studied the target intervention of miners' unsafe behavior, and the research results have practical significance. The method and process can also be used to intervene on the unsafe behavior of workers in other industries. It is necessary to redefine the data boundary according to the behavior characteristics of workers in different industries, and make adaptive adjustments to the node location process by combining the sources of unsafe behavior data and the behavior characteristics of workers in various industries. This is a possible direction of future research and development.

**Author Contributions:** R.T. conceived the key ideas and the paper structure; Y.Y. and X.M. conducted the data research; Y.Z. and Q.J. discussed the results of data analysis; R.T. and G.S. wrote the paper; R.T. and Y.Z. reviewed the process.

**Funding:** The study was financially supported by the National Natural Science Foundation of China (No. 51674268).

**Conflicts of Interest:** The authors declare no conflict of interest.
