The study of sector complexity can effectively improve the accuracy of sector dynamic capacity assessment and provide reference for air traffic decision-making. This paper firstly establishes the approach and departure procedures’ potential exceeding conflict complexity metrics and key conflict node complexity metrics to quantify the impacts of different dynamic traffic indicators; secondly, using machine learning to train and test the different features and the number of passing aircraft in the sector and get the mapping relationship, the model can be used to predict the sector traffic capacity under different traffic flow configurations; finally, uses the one-week operation data of AP16 from complex terminal for example verification, and the result shows that the machine model can be predicted the sector traffic capacity. The results show different traffic configurations would have an impact on sector capacity, and the dynamic capacity of the sector in the nearly balanced phase of approach and departure reached a maximum of 15 flights/15min. In addition, this paper applies Shapley"s additive interpretation to quantify the contribution of each feature to the predicted sector capacity, which can provide a reference for the subsequent planning and design of the sector.