Research On Airport Group Delay Prediction Based On Skip-LSTM

With the rapid development of the air transport industry,many airport groups have been formed at home and abroad,and the development of airport groups has become increasingly mature.At the same time,with the development of big data and deep learning technologies,it has been widely used in various fields.Therefore,a method based on Skip Long Short Term Memory(Skip Long Short Term Memory,Skip-LSTM)for airport group delay prediction is proposed in this paper taking airport group delay prediction as the goal,and based on big data and deep learning technology.The purpose is to use the deep learning algorithm to mine the data of the Beijing-Tianjin-Hebei Airport Group and the New York Airport Group in the United States to predict the delay situation of the China-US airport group and compare the differences between the China-US airport group data set.Provide auxiliary reference opinions for decision-making of airports,air traffic management bureaus,airlines and other related departments.The main research work of the paper is as follows:Firstly,an airport group data set is constructed and the data is preprocessed: flight data and meteorological data of each airport in the target airport group are extracted,and flight data and weather data are fused using time as a key value.Finally,the data of each airport in the airport group is merged.the construction of the airport group data set is completed;then the data is encoded according to the characteristics of each feature in the airport group data.Different coding methods for different airport group data and different data features are adopted in this paper,so that the feature information in the airport group data set can be more accurately extracted by the deep learning network and better prediction result is obtained.Secondly,an airport group delay prediction method based on Long Short Term Memory(LSTM)is proposed.Because the delay status of the airport group is time series,therefore,a recurrent neural network(RNN)with time series is used to delay the airport group.However,the traditional RNN has the problem of gradient disappearance or gradient explosion during the training process,and it is impossible to learn long-term sequences.The forgetting gate,input gate,and the output gate are added in the airport group delay prediction model based on the LSTM network which can avoid the disappearance and explosion of the gradient to a certain extent,and longer time series information can be learned and higher accuracy is obtained.Finally,because the LSTM needs to update all the neuron states at each moment,the training time is longer and the model prediction accuracy cannot continue to improve.Therefore,an airport group delay prediction model based on Skip-LSTM is proposed in this paper.A skip gate is added in this model to the LSTM.The skip gate controls the update of neuron states,so that the time correlation of the delay status of the airport group is more fully extracted,and a higher accuracy rate and a shorter training time are obtained.Finally,the softmax classifier is used in the model to classify and predict the delay status of the airport group.The final model's prediction accuracy rate is 95.35%,which is superior to the traditional network model and can effectively predict the delay status of the airport group.