The theory of nu-support vector regression (Nu-SVR) is employed in modeling time series variation for prediction. In order to avoid prediction performance degradation caused by improper parameters, the method of parallel multidimensional step search (PMSS) is proposed for users to select best parameters in training support vector machine to get a prediction model. A series of tests are performed to evaluate the modeling mechanism and prediction results indicate that Nu-SVR models can reflect the variation tendency of time series with low prediction error on both familiar and unfamiliar data. Statistical analysis is also employed to verify the optimization performance of PMSS algorithm and comparative results indicate that training error can take the minimum over the interval around planar data point corresponding to selected parameters. Moreover, the introduction of parallelization can remarkably speed up the optimizing procedure.