The prediction of MHC-II binding peptides has long been a principal challenge in immunology. Recently, the modeling of MHC-II binding peptides has come to emphasize quantitative prediction, instead of categorizing peptides as no-binder or high-binderand moderate-binder. In this paper, we develop a support vector machine regression’s (SVR) approach to predict MHC-II binding peptides. Considering global description of the peptide sequences, input vectors of same lengths are generated from peptides of different lengths, and then support vector machine regression is used to model binding affinities between MHC-II molecules and peptides; at last we obtain the prediction model called SVRMHC-II. When applied to three MHC-II alleles, SVRMHC-II produces better predictions than several prominent methods in terms of area under ROC curve, indicating it is an effective tool.