biostudies-literatureLiu KNatural Science Basic Research Program of Shaanxi ProvinceNational Natural Science Foundation of ChinaSpecial Fund for the Launch of Scientific Research in Xijing University27327https://www.ebi.ac.uk/biostudies/studies/S-EPMC12301436biostudies-literatureUnknown15(1)To effectively identify the source of water in coal mines and prevent water-related accidents, this paper utilises the hydrochemical characteristics of the aquifers Shanxi Hanzui Coal Mine. The fuzzy C-means (FCM) clustering method is employed to classify water sample data, followed by principal component analysis (PCA) for dimensionality reduction to extract key features. The SMOTE algorithm is then applied to address the issue of class imbalance. Based on this, a decision tree model (FPS-DT) is constructed using the CART algorithm. To validate the model's performance, five-fold cross-validation was used for evaluation. The results showed that the average classification accuracy of the FPS-DT model was 93%. In contrast, the accuracy of the comparison model, which only used PCA and decisioScientific reportsIdentification of water sources of mine water bursts based on the FPS-DT model.PMC123014362025JC-BMS-5112021JLM-48XJ18T0452104222XJ24B1251909224Xia YLi XLiu KWang QZhao SSun CYue WChen SfalseIdentification of water sources of mine water bursts based on the FPS-DT model.To effectively identify the source of water in coal mines and prevent water-related accidents, this paper utilises the hydrochemical characteristics of the aquifers Shanxi Hanzui Coal Mine. The fuzzy C-means (FCM) clustering method is employed to classify water sample data, followed by principal component analysis (PCA) for dimensionality reduction to extract key features. The SMOTE algorithm is then applied to address the issue of class imbalance. Based on this, a decision tree model (FPS-DT) is constructed using the CART algorithm. To validate the model's performance, five-fold cross-validation was used for evaluation. The results showed that the average classification accuracy of the FPS-DT model was 93%. In contrast, the accuracy of the comparison model, which only used PCA and decisio2025-01-01T00:00:00Z2025 Jul2026-03-31T11:05:41.809Z2025-08-28T03:07:58.687ZS-EPMC123014364071717610.1038/s41598-025-13301-y