Asphalt pavement performance prediction is complex and nonlinear when it involves multi-factor. In order to overcome the defects existing in traditional prediction models, a long-period and multi-factor prediction model with high precision needs to be established, on which the dimension of each factor is reduced by grey relational analysis, and the important relational factors are selected for nonlinear prediction by support vector machine regression. Accordingly the performance prediction model of asphalt pavement based on GRA-SVR was proposed and the measured RDI from Guangyun freeway were collected as an example to validate the proposed model. The results show that GRA-SVR model has better accuracy and maneuverability compared with GM(1,1)and PPI models. It can be used in long-term process and provide model reference for large data maintenance decision-making.