Predictive Modelling: Flight Delays and Associated Factors, Hartsfield–Jackson Atlanta International Airport

Abstract

Nowadays, a downside to traveling is the delays that are constantly being advertised to passengers resulting in a decrease in customer satisfaction and causing costs. Consequently, there is a need to anticipate and mitigate the existence of delays helping airlines and airports improving their performance or even take consumer-oriented measures that can undo or attenuate the effect that these delays have on their passengers. This study has as main objective to predict the occurrence of delays in arrivals at the international airport of Hartsfield-Jackson. A Knowledge Discovery Database (KDD) methodology was followed, and several Data Mining techniques were applied. Historical data of the flight and weather, information of the airplane and propagation of the delay were gathered to train the model. To overcome the problem of unbalanced datasets, we applied different sampling techniques. To predict delays in individual flights we used Decision Trees, Random Forest and Multilayer Perceptron. Finally, each model’s performance was evaluated and compared. The best model proved to be the Multilayer Perceptron with 85% of accuracy.