Abstract：Aiming at the recognition problem of mine groundwater mixed with high degree of water inrush source，using the multivariate statistical analysis principle and the mixed calculation principle comprehensively，this paper established the identification model and mixed model of a mine water inrush, based on actual data as training samples, analysis and test on them respectively. The results show that logistic analysis can effectively establish a low degree of mixed water inrush identification model with lower error rate and the mixed model uses the principal component analysis. Many indexes affect the safety of mine water inrush which result in a high dimensional and sparse safety evaluation matrix, while the surveillance data provided by the limited time and space are incomplete. A SVM model based on the related relation was presented. The correlation of indexes and the error sums of each indexes were calculated, and were used to rank indexes. SVM is reused to achieve ultimate small sample evaluation based on the new indexes with maximum accuracy rate. The analysis results show that the correlation is high among Ca2+，Mg2+，K++Na+；the correlation is lower between K++Na+and CL-, and also between HCO3- and SO42-; Ca2+ is more important and SO42-和CL- is less important; The recognition result is relatively high and cost low without redundant information.