The human error of air traffic controller is the main cause of the aircraft collision in the air, and it is a difficult problem to effectively predict the failure probability of human factor of Air Traffic Control (ATC). For this, on the basis of Technique of Human Error Rate Prediction (THERP) and Cognitive Reliability and Error Analysis Method (CREAM), a hybrid prediction method is proposed for failure probability of ATC human factor. When using THERP to evaluate the complex behaviors of controllers, it is difficult to accurately quantify various effects with fewer available data and rough correction calculations. CREAM is introduced to make up for the shortcomings of THERP by quantifying human behavior factors from the early stage and incorporating them into the background reasons for comprehensive consideration, in order to improve the prediction reliability for failure probability of ATC human factor. The actual example shows that the method can not only carry out a more scientific and reasonable analysis, but also find the weak links to formulate targeted control measures. This provides an effective analysis method for studying the failure probability prediction of ATC human factor.