Based on Gaussian process (GP), a new parameters’ correlation analysis method for injection molding is proposed. Referred to the design idea of canonical correlation analysis (CCA), GP is used to extract accurate canonical correlation variables simultaneously from two data sets. And then the canonical correlation variables are used to analyze the correlation between parameters and design objectives. The cross member under windshield of a van is taken for a case. For the weld lines defects produced in injection process, the correlation of process parameters is analyzed to identify which parameters are more related to weld lines. The validity of this method is proved by the optimal result. And this provides strong theory and feasible algorithm for adaptive intelligent optimization and controlling of the parameters in injection process.