Risk assessment of floor water inrush using random forest, VIKOR, and GIS: A case study

As mining depth continues to increase, coal mines face increasingly complex water inrush conditions and varying risk factors, making the assessment of floor water inrush (FWI) risk more challenging. To more accurately evaluate the risk of FWI, this study proposes a composite evaluation model integrating Random Forest (RF), the VIKOR method, and Geographic Information System (GIS). Taking the Yangcheng Coal Mine in North China as the study area, an evaluation index system for FWI risk is established, consisting of six key factors: fault fractal dimension (C1), seam dip angle (C2), mining depth (C3), key-strata thickness (C4), water pressure (C5), and dip length of the working face (C6). The RF algorithm is used to mine the data features, ensuring a more scientific and reasonable calculation of the weights. The results show the following weights: C1 = 0.108, C2 = 0.092, C3 = 0.168, C4 = 0.171, C5 = 0.368, and C6 = 0.093. GIS technology is employed to collect quantitative data of six key factors affecting FWI at 1314 grid center points in the mining area. The VIKOR method is then applied to process the data, obtaining a risk ranking for each point, with higher-risk points ranked first. The evaluation results are visually presented in GIS with a color gradient, transitioning from green to red, indicating increasing risks. Actual water outflow locations are found within the red zones, validating the effectiveness of the model. This FWI risk assessment method provides new insights and practical references for the prevention and control of water-related hazards in coal mining.