This study concerns the combination of Raman spectroscopy and multivariate statistical analyses for the assessment of lymph nodes in the course of breast cancer diagnostics and staging. Axillary lymph node samples derived from breast cancer patients were measured by Raman microspectroscopy. The resulting Raman maps were pre-processed and cleaned of background noise and low intensity spectra using a novel method based on selecting spectra depending on the distribution of the mean of arbitrary units of all spectra within individual samples. The obtained dataset was used to build different types of Support Vector Machine (SVM) models, including linear, polynomial and radial basis function (RBF). All trained models were tested with an unseen independent dataset in order to allow an assessment of the predictive power of the algorithms. The best performance was achieved by the RBF SVM model, which classified 100% of the independent testing data correctly. In order to compare the SVM performance with traditional chemometric methods a linear discriminant analysis (LDA) model and a partial least square discriminant analysis (PLS-DA) model were generated. The results demonstrate the enhanced performance and clinical potential of the combination of SVMs and Raman spectroscopy and the benefits of the implemented filtering.