Effectively identifying the main influencing factors of remote tower controllers' situation awareness (SA) level can better provide a reference basis for remote tower control design and use. First, data collection tests were conducted on controllers in traditional and remote tower environments to analyze the differences of subjective and objective indicators in the two environments. Secondly, hypothesis testing was used to verify the feasibility of eye movement metrics as an evaluation of SA for remote tower controllers. Then, a combination of K-means clustering and support vector machine (SVM) modeling analysis was used to classify and identify the SA level of remote tower controllers based on the sensitive eye movement indicators. The results showed that the accuracy of SA recognition reached 99.72% by using Poly kernel function to train the model. The results confirm that the combined model of K-means clustering and support vector machine model can be used as an effective method to analyze the SA of remote tower controllers.