Xishan gold mine is subordinate to the Sanshandao gold mine and located in the coastal area of Laizhou Bay，Laizhou City，Shandong Province.In terms of geotectonic，it is located in the western part of the second up-warping zone of the Neocathaysian structural system，which is also Sanshandao-Cangshang fracture of the eastern side of Yishu deep facture.Xishan gold mine has been exploited in under the Bohai sea.Submarine mine water inrush has become an urgent problem to be solved in mine mining.The research on subway water can classify the types of mine groundwater and then predict the possibility of water inrush.Taking groundwater system of Sanshandao gold mine for example，hydrochemical data of 31 water samples was chosen to study with multivariate statistical analysis methods.By using factor analysis，it can reduce the spatial dimension of many variables with correlation relationship，and then identify principle factors which represent over ninety percent information of hydrochemical data.Hierarchical clustering analysis（HCA）uses these principle factors as clustering variables.HCA combined with actual groundwater quality divided the studied groundwater into 2 classic groups，then established and validated Fisher identification model.Through FA and HCA，the groundwater of -375 m subway were divided into two types which all have a specific discriminant function could determine which type of water is.The results represent that the water samples were divided into two typical M1 and M2 by factor analysis combined with principle component analysis.Among the 31 water samples，three of them were discriminated wrong，and the correct rate of discriminant reached 90.3%.Stepwise discriminant analysis and factor analysis were combined to process the seven conventional ions data.Bayes linear discriminant function and function values from 1740 exploration line to 2740 exploration line in -375 m sublevel was obtained.Bayes linear function discriminant results are completely consistent with the results of the factor analysis method，and the two selected discriminant water samples also agree.The consistency of the discriminant results shows that the factor analysis method and the stepwise analysis method are mutually verified.A multivariate statistical method was combined to obtain a quantitative Bayes linear discriminant function，which was applied to the recognition of the source type in the mining area.It was only necessary to know the ion concentration of the corresponding variable，and the water sample type could be determined by substituting it.This method has the characters of accurate，fast，and economical.