Flight delay predictions and the study of its causal factors using machine learning algorithms

Abstract

In recent years, the global civil aviation industry has been developing rapidly. Due to the rising demand for air transportation, airports are confronting saturation problems. Heavy traffic and long queues are expected for take-off and landing. Hence, the physical constraints have magnified the problem of having surging flight delays. Yet, the operational efficiency and the reputation of the airport will deteriorate if the delay propagates. Additional expenditures are also expected. Several machine learning approaches were adopted in this research to predict flight delay, including the decision tree, random forest, k-nearest neighbour, Naïve Bayes, and artificial neural networks. The results show that all algorithms achieved more than 80% of accuracy and artificial neural networks perform the best among the alternatives. While Naïve Bayes is the least accurate, k-nearest neighbour have the lowest F <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf> score.