The Hilbert-Huang （HHT） boundary problem appeared when the signal is decomposed by empirical mode decomposition method (EMD) and the intrinsic mode functions (IMF) in Hilbert transform. In order to overcome the problem, the HHT boundary optimization method based on discrete uniform immune algorithm (DUIA) and support vector regression (SVR) is proposed in this paper. To effectively analyze the boundary problem of HHT, this scheme can use DUIA to optimize parameters of SVR, and then predict the signal by the trained optimally SVR model. For the sine superposition and practical signals, the corresponding simulation results demonstrate that the proposed algorithm can solve the boundary problem of HHT effectively, and its performance is better than prediction method by SVR.