Groundwater quality in arid/semi-arid coal-mining areas is increasingly threatened by intensive human activities. However, quantitatively linking specific pollution sources to health risks and translating this into risk-based management priorities remains a critical challenge. This study addresses this gap by proposing an integrated framework combining self-organizing maps (SOM), positive matrix factorization (PMF), and Monte Carlo simulation to systematically decipher groundwater contamination sources and prioritize associated health risks, using the Huolingol mining area in Inner Mongolia, China, as a representative case. SOM analysis recognized three hydrochemical clusters: Cluster 1 (Na-HCO3/Ca-Na-HCO3 type), influenced by cation exchange and silicate weathering under alkaline conditions, exhibited elevated F- and NH4+-N concentrations. Cluster 2 (Ca-Na-SO4/Ca-Na-HCO3 type) was characterized by elevated TDS, TH, SO42-, and NO3-, indicating significant anthropogenic impact. Cluster 3 (Ca-HCO3 type) represented the natural background with minimal contamination. Source apportionment using PMF revealed that human-driven composite pollution is the main cause of water quality evolution, with the mixed source of coal mine drainage and industrial/domestic wastewater contributing the highest proportion (22.42%), highlighting the overwhelming impact of mining and industrial activities. Monte Carlo simulation showed children faced a 33.9% probability of exceeding the safe health threshold (HI > 1), 1.75 times higher than adults (19.4%). F- and SO42- were the main risk drivers. Source-specific analysis indicated geogenic fluoride (Factor 1) contributes similar to 48% of total health risk, followed by organic degradation (similar to 26%) and anthropogenic sulfate (similar to 23%). This framework quantitatively links sources to health risks and provides actionable guidance for groundwater management in water-stressed mining regions.