Abstract: An automated traffic classification system is proposed to solve the problem of machine learning in network traffic analysis tasks. By combining semantic and binary flow representation techniques, it generated a unified representation of network traffic data packets, which was then applied to feature representation and model training. This network traffic representation method was integrated with automated machine learning to establish a compatible and general automated machine learning flow representation and classification system. This approach significantly reduced the need for feature extraction and model tuning in various traffic analysis tasks, thereby facilitating the broader application of machine learning techniques in traffic analysis. The proposed system was evaluated using the ISCX2016-VPN and Kitsune datasets. Experimental results show that this system performs well on these datasets.