In the process of mine water inrush disaster prevention and control,it is very important to accurately and quickly identify the type of water inrush sources for coal mine safety production. However,the traditional hydrochemical method has the disadvantages of time-consuming and complex detection. Therefore,a new idea of identifying mine water inrush sources using Raman spectroscopy is proposed. First of all,the water samples of goaf water,roof sandstone fissure water,Ordovician limestone water,Taiyuan limestone water,surface water and their mixture were collected from the Huainan mining area as experimental objects,and the Raman spectral data of water samples were collected with the help of Raman spectroscopy system. Then,the common spectral pretreatment method is used to reduce the noise of the original Raman spectrum. Then,the whale optimization algorithm (WOA) is used to screen the characteristic information of the water sample,and the characteristic information that best represents the mine water sample is obtained. Finally,the filtered characteristic Raman information is used as input to construct BP neural network (BPNN),K-nearest neighbor algorithm (KNN),support vector machine (SVM),decision tree (DT) and naive Bayesian (NB) classification models respectively,to verify the feasibility of Raman spectrum combined with WOA screening characteristic Raman information for mine water source identification. Experiments show that 102 characteristic Raman information can be filtered from 2 048 Raman data points by WOA,reducing the number of Raman information to 4.98%,and the modeling accuracy of characteristic Raman information filtered by WOA is higher than that of full Raman data. In addition,when the characteristic Raman information filtered by WOA is used to build BPNN,KNN,SVM,DT, and NB water source identification models,the analysis speed has been improved to varying degrees. The research results show that using WOA to screen the characteristic information of the Raman spectrum of the mine water source can effectively reduce the redundancy of the Raman spectrum data and significantly improve the speed of the Raman spectrum analysis,which can provide a reference for the rapid detection of the mine water source.