ATC’s task is to ensure safety of the aircraft, it is necessary to run air traffic control system safety assessment and early warning to ensure flight safety. It establishes ATC operation state vector space with starting from the main risk factors influencing running safety based on ATC big data concepts .Using principal component analysis (PCA) to remove the basic vector linear correlation and using clustering analysis to establish the ATC safety warning knowledge base, we can realize the ATC safety assessment and early warning in time. Examples show that the early warning classification analysis method can achieve good classification and safety assessment in the actual operation, it also can provide reference for control department daily decisions.