In view of the shortage of the empirical analysis which can be used to evaluate the performance of the modified asphalt mixture, the K means clustering algorithm was introduced considering the characteristics of the highway project construction. The pavement performance data of typical modified asphalt for highway and laboratory test data were thoroughly searched. The data were iteratively analyzed by the K means clustering algorithm. Based on the study, the pavement performance classification standard of modified asphalt mixture was put forward. In the standard the dynamic stability is used as an indicator to evaluate high temperature performance, grading standards for dense graded asphalt mixture in surface layer are 3200, 5000, 8000 times/mm，those in middle and lower layer are 3200, 4500, 6800 times/mm, and grading standards for SMA are 3600, 6000, 8000 times/mm. The classification standards for moisture susceptibility are 81%，87%，93% in terms of freeze thaw splitting strength and are 84%，89%，94% in terms of residual stability. In terms of bending failure strain, the classification standards for low temperature performance are 3000×10-6，3500×10-6，5000×10-6.