This paper use variance reciprocal method to determine the weight, and combine a linear regression model, the grey forecasting model with exponential smoothing model. That purpose is that we can forecast the processing tomato yield accurately. We build a processing tomato production combination forecasting model by the 2001-2009 Xinjiang a processing tomato planting and production data. Take the predicting yield of 2010 and 2011 tomato production for example, we proved that compared with the single forecast model, the combination forecasting model was more accurate and reliable. And it provided the foundation for the establishment of the processing tomato industry decision support systems.