The domestic patent applications by region (DPAR) are complex to conduct due to its nonlinearity of influenced factors. It is necessary to make a trade off among these factors when some of them conflict. A novel way about nonlinear regression modeling of DPAR with the potential support vector machines (P-SVM) is presented in this paper. In the model development, a simulated annealing (SA) algorithm is employed to optimize P-SVM parameters selection. Using a set of real data from China, simulation results show that the SA optimization algorithm is robust. The results also indicate that methods based on the P-SVM perform are better than the ones based on the standard SVM in both model generalization and DPAR forecasting accuracy.