In road engineering, UV radiation accelerates chemical oxidation and performance degradation during long-term service. This study proposed a machine learning-based framework for intelligent identification of UV aging states in asphalt using multi-source microscopic features. SBS modified asphalt was subjected to different UV aging durations. The conventional performance, rheological behavior, and microscopic characterization tests were used to investigate the performance of UV aged asphalt. Based on the Pearson correlation analysis, the four representative microscopic parameters, including carbonyl index (I_C=O), sulfoxide index (I_S=O), polydispersity index (PDI), and large molecular size fraction (LMS), were selected to construct a multi-source microscopic feature set. Subsequently, a linear Support Vector Machine (SVM) model combined with Leave-One-Out Cross-Validation (LOOCV) was developed for UV aging state classification. The results showed that the proposed model achieved an accuracy and balanced accuracy of 83.3%, with a macro-F1 score of 82.2%. Feature contribution analysis indicated that LMS and I_S=O were the dominant variables for aging state identification, with a cumulative contribution of 62.6%. Furthermore, the predicted aging categories exhibited consistent trends with experimentally measured performance degradation. The proposed framework establishes the relationship among microscopic structural features, aging states, and service performance deterioration, providing a physically interpretable data-driven approach for intelligent evaluation of asphalt UV aging states.