Non-tuberculosis mycobacteria (NTM) are the collection of mycobacteria other than Mycobacterium tuberculosis complex (MTC) and Mycobacterium leprosy. The clinical symptoms of NTM are very similar to MTC infection, yet their treatments are different, thus rapid and accurate identification methods of NTM are urgently needed. Single-cell Raman Spectroscopy (SCRS) is label-free, and independent of cultivation, thus it is deemed a rapid and efficient technology with low cost. Here we propose an SCRS based method to identify NTM based on confocal SCRS. We selected six common NTM species in the clinic, Mycobacterium abscessus, Mycobacterium gordonae, Mycobacterium fortuitum, Mycobacterium fortuitum, Mycobacterium avium and Mycobacterium kansasii. The unsupervised low-dimensional visualization t-distribution random neighborhood embedding method for the data structures proved the separability of data in the low-dimensional space. Performance of six commonly classifiers, including Support Vector Machine (SVM), K-Nearest Neighbor method (KNN), Partial Least Square-Discriminate Analysis (PLS-DA), Random Forests (RF), Linear Discriminant Analysis (LDA) and XG Boost was compared, with SVM and LDA achieving an accuracy of 99.4% and 98.8% respectively in NTMs classification. SVM offers 100% classification accuracy for every species, except Mycobacterium kansasii which is slightly lower (97.96%, 48/49), while LDA offers 100% accuracy for each species except Mycobacterium abscessus (95.65%; 22/23) and Mycobacterium gordonae(96.30%, 26/27). Therefore, SCRS combined with SVM can accurately classify NTMs and thus provide a new tool for the rapid diagnosis of NTM.