Accurate flight ground service time prediction can improve the flight transit efficiency and realize flight refinement management. However， the lack and abnormality of relevant data make the research more challenging in real scenarios. To this end， a flight ground service time prediction model considering missing values is proposed. A dynamic time warping（DTW） algorithm is introduced on the basis of the causal graph convolutional network （CGCN） to realize the prediction of flight ground service link time under different data missing modes and missing rates. The flight support dataset （6 480 items） of a large airport in China is used as an example for validation. The results show that the proposed model can maintain high prediction performance under conditions of 20%—80% missing rates， compared with the remaining seven benchmark models including causal graph convolutional network with missing data （CGCNM）， dynamic spatial-temporal graph convolution network （DSTGCN）， Bayesian temporal matrix factorization （BTMF）， long short-term memory （LSTM）， etc. The mean absolute error （MAE） of the prediction results for each service time node is reduced by more than 8.1%， and the root-mean-square error （RMSE） is reduced by more than 4.6%. The experiment demonstrates that the proposed model is better than the baseline model in terms of prediction accuracy and prediction stability. It can provide an objective and reliable decision-making basis for flight support operations.