The Verkhnekamsk potassium-magnesium salt deposit (VKSD) is one of the largest in the world. The primary challenge in underground salt mining is maintaining the integrity of the groundwater protective layer, which separates the mined seams from aquifers. In this context, the Verkhnekamsk deposit is mined using a chamber system room-and-pillar method, ensuring the stability of the protective layer through inter-chamber pillars. This paper presents the results of a preliminary analysis of the geological and mining conditions in one of the mines of the Verkhnekamsk deposit. The procedure for forming the initial data set is discussed. Test calculations based on a limited data set were performed, demonstrating the potential of combining artificial neural network algorithms and discrete mathematical analysis. The results achieved on the formed dataset successfully identified hazard classes. Thus, it can be concluded that this technology is fundamentally effective for assessing the risk of groundwater protective layer failure. The proposed approach establishes links between phenomena, their associated risks, and the deformations of underground workings and the Earth's surface, enabling proactive measures to protect mines from flooding.