Flight Delay Propagation Prediction Model Based on CNN-MogrifierLSTM with Attention Mechanism SimAM

‍ ‍Large scale flight delays are often caused by the spread of delay propagation in continuous flights. The effective reference of the civil aviation departments is presented and the related economy is reduced by predicting the flight delay in advance. Firstly， datasets are to be cleaned and fused in this model. Two different construction methods for strong spatial flight chain datasest and strong sequential flight chain datasets are designed according to the actual flight operation situation of air traffic control department. Secondly， CNN-MogrifierLSTM with attention mechanism SimAM is proposed based on the space-time characteristics of the flight delay spread， SimAM-CNN module is used for primary extraction of spatial features， and then MogrifierLSTM network is applied for learning time information. Finally， softmax classifier is utilized to classify and predict the delay levels of flights. The prediction method in this paper achieved an accuracy of 93.16%， which is better than either CNN or LSTM alone. With the addition of SimAM attention mechanism， the accuracy of CNN-MogrifierLSTM network on different datasets are also improved by about 0.6%.