Abstract: In order to solve the problem that traditional pavement performance evaluation method is subjective and the existing models have defects, the evaluation index system of freeway asphalt pavement performance was analyzed.The number of evaluation index was reduced by the principal component analysis (PCA) method to form independent principle components, then the principle components samples were selected to train the support vector machine (SVM). The PCA-SVM model combining principal component analysis and support vector machine was established to evaluate freeway asphalt pavement performance. Finally, the pavement performance data collected from Luoning freeway in 2016 was used to verify the proposed model, and the results were compared with "Highway performance assessment standards". The results show that the evaluation results of PCA-SVM model are in accordance with "Highway performance assessment standards", but the results of the 5 sections with higher repair rate are one grade lower than that of "Highway performance assessment standards".The method provides a reference for the improvement of highway performance assessment.