Density analysis based on flight delay prediction with genetic algorithm hyperparameter tuning

Abstract

The aviation sectors have been growing continuously. Flight delays are a major problem in the current aviation system. Once there is a flight delay, this causes chain delays at multiple airports, which results in tremendous economic loss. To deal with the delay and early arriving flights, flight delays prediction models were proposed by many researchers in order to prevent and avoid these problems. In this study, machine learning approaches for the flight delays prediction were proposed to predict the arrival delay in the next 15-minute interval window. Three different models of LSTM, Random Forest and XGBoost were utilized with the data of the flight in the United States in year 2015. The genetic algorithm was applied as a tuning parameter for optimization to the improvement of the prediction performance. The models were validated and compared to the model efficiency though statistical indicator. The satisfactory results with coefficient of determination of 0.6659, 0.6713 and 0.6760 were obtained from LSTM, Random forest and XGBoost respectively. The application of methods in which historical data were utilized showed promise for the improvement of arrival delay in the next 15-minute interval window prediction capabilities