Abstract: Through the investigation and collection of hydrogeological data in the Gubei Coal Mine, by using SPSS factor analysis method, various hydrochemical ion indexes were concentrated, extracted and synthesized into principal factors, to use a few variables to reflect the completeness of information to the greatest extent. The factors suitable for model discrimination were applied to the Bayesian multi-class linear identification model, and the samples of sandstone fractured water and limestone water were distinguished respectively. The results show that the comprehensive discrimination accuracy rate of the sandstone fractured water and limestone water is 86.9%. Compared with the direct Bayesian discrimination, the complexity of the analysis process is decreased and the discrimination is more accurate.