Abstract: Existing studies indicate that the longer the queue length of vessels, the greater the channel saturation. To predict channel congestion, this study proposes a congestion prediction method that considers the maximum queue length based on the fundamental principles of traffic wave theory. The model utilizes Automatic Identification System（AIS） data to extract traffic flow characteristic parameters and, considering the differences in navigation behavior among ships in different waters, proposes a method for dividing the channel into characteristic areas. The queue length in traffic wave theory is selected as the evaluation index for congestion, and a method for predicting the maximum queue length based on Gaussian process regression is proposed to achieve the prediction of waterway congestion levels. A case study is conducted in the Yuxi River section of the Yangtze River Basin. The results show that the theoretical value of the maximum queue length in this section in July 2020 is 0.98 km, and the Adjusted R² index of the established regression model is 0.88, predicting a maximum queue length of 1.34 km with an error of 0.37 km compared to the theoretical value. The research results demonstrate that the proposed model has a high degree of interpretability and can effectively predict the maximum queue length, thereby enabling the prediction of channel congestion. This study provides a theoretical basis for improving the level of maritime supervision services.