A total of 10 quaternary loose rock pore water samples were collected from Suizhong County, Liaoning. The pH, Cl^-, SO₄^2-, NH₄⁺, NO₂^-, NO₃^-, F^-, total hardness, total dissolved solids, iron, manganese, zinc, cyanide and volatile phenols were considered as the water quality parameters. Rough set theory was employed for data reduction. Meanwhile, to find attribute reduction set, the attribute dependence degree and information entropy heuristic algorithms were combined. Support vector machine was employed to evaluate groundwater quality for all parameters before and after reduction, respectively. The results showed that rough set theory reduced the number of chemical parameters from 14 to 8, and assessment results with attribute reduction were the same as those without attribute reduction. The groundwater quality in the study area was mainly class II and III, which meets the permissible limits. However, iron and three nitrogen were exceeded drinking water quality standard. Although the combination of rough set and support vector machine reduced redundant indices, the accuracy of water quality classification remained effective, while the complexity of calculation was reduced and the rationality of assessment results was guaranteed.