Departure Flight Delay Prediction and Visual Analysis Based on Machine Learning

Nowadays, the rapid growth of civil aviation transportation demand has led to more frequent flight delays. The major problem of flight delays is restricting the development of municipal airports. To further improve passenger satisfaction, and reduce economic losses caused by flight delays, environmental pollution and many other adverse consequences, three machine learning algorithms are constructed in current study: random forest (RF), gradient boosting decision tree (GBDT) and BP neural network (BPNN). The departure flight delay prediction model uses the actual data set of domestic flights in the United States to simulate and verify the performance and accuracy of the three models. This model combines the visual analysis system to show the density of departure flight delays between different airports. Firstly, the data set is reprocessed, and the main factors leading to flight delays are selected as sample attributes by principal component analysis. Secondly, the mean absolute error (MAE), mean absolute percentage error (MAPE) and root mean square error (RMSE) were selected as evaluation indexes to compare the prediction results of three different models. The final results show that the departure flight delay prediction model based on BPNN algorithm has faster solution speed and overcomes the over-fitting problem, and has higher prediction accuracy and robustness. Based on the algorithm developed in this paper, the airport system can be planned in a targeted manner, thereby alleviating the pressure of air transportation and reducing flight delays.