Based on the monitoring data of February~December 2009 from Xinxue River Constructed Wetland, the regression analysis theory was used to establish mathematical models to determine the correlationship among the removal rate of pollutants, pollutant concentration, water temperature and influent water inflow, and optimize the operation parameters of the wetland. The multivariate linear regression equation was established with Statistical analysis software (SPSS). According to the correlation coefficient of the equation: the NH3-N removal has a significant positive correlation with water temperature, and a significant negative correlation with influent flow, total nitrogen removal has a significant positive correlation with water temperature, COD removal has a significant positive correlation with COD influent concentration. In addition, veracity of the model was also checked by contrasting the measured and analog data of March~June 2010 from the wetland. The results show that this model has a desired effect to estimate the removal rate and effluent concentration of COD and NH3-N, but undesirable to total nitrogen.