Prediction of weather-induced airline delays based on machine learning algorithms

• Published in: IEEE/AIAA Digital Avionics Systems Conference pp. 1 - 6
• Main Authors: Sun Choi, Young Jin Kim, Briceno, Simon, Mavris, Dimitri
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.09.2016
• Subjects: Atmospheric modeling; Data models; Delays; Meteorology; Predictive models; Schedules; Training
• ISSN: 2155-7209
• DOI: 10.1109/DASC.2016.7777956

The primary goal of the model proposed in this paper is to predict airline delays caused by inclement weather conditions using data mining and supervised machine learning algorithms. US domestic flight data and the weather data from 2005 to 2015 were extracted and used to train the model. To overcome the effects of imbalanced training data, sampling techniques are applied. Decision trees, random forest, the AdaBoost and the k-Nearest-Neighbors were implemented to build models which can predict delays of individual flights. Then, each of the algorithms' prediction accuracy and the receiver operating characteristic (ROC) curve were compared. In the prediction step, flight schedule and weather forecast were gathered and fed into the model. Using those data, the trained model performed a binary classification to predicted whether a scheduled flight will be delayed or on-time.