Construction of a Large-scale Flight Schedule Prediction Model Based on Autoregressive Method

With the rapid development of civil aviation industry and the growing contradiction between demand of passenger and cargo transportation and civil aviation comprehensive service ability, flight punctuality rate is under high pressure. It is necessary to accurately predict the actual flight take-off and landing time to better meet passenger transportation needs and improve the efficiency of civil aviation operation. This paper studies the time series characteristics of civil aviation flight and the structural characteristics of flight operation, constructs an autoregression model with the input of history delay data and the output of flight delay on predicted day, optimizes the model by changing the algorithm of matrix and vector, finally reduces the calculation time by $\mathbf{80 \%}$ and improves the prediction accuracy by $43 \%$ compared to traditional models directly applying deep learning.