In order to study the relationship between mixture gradation variation and road performance, the road performance of asphalt mixture was predicted. In this paper, a typical asphalt mixture grading variation model was constructed, and the screening residual characteristic parameters were proposed to characterize the mixture grading variation characteristics. SPSS software was used to analyze the correlation between the structural parameters of the mixture and the road performance, and a prediction method for the road performance of asphalt mixture was established, with the asphalt mixture structural parameters as the medium. Through regression analysis, the relationship equation between the variable gradation screen residual characteristic parameters and the road performance index is established, and the qualification criteria of each road performance index value is determined. The results show that the prediction of Marshall stability, freeze-thaw splitting strength ratio, creep rate and residual stability of asphalt mixture are in good agreement with the actual results. It can be used to predict the partial pavement performance of asphalt mixture with variable gradation, adjust the construction mix quickly without affecting the construction process, and ensure the pavement performance of asphalt mixture.