Flight arrival delay prediction using gradient boosting classifier

Abstract

The basic objective of the proposed work is to analyse arrival delay of the flights using data mining and four supervised machine learning algorithms: random forest, Support Vector Machine (SVM), Gradient Boosting Classifier (GBC) and k-nearest neighbour algorithm, and compare their performances to obtain the best performing classifier. To train each predictive model, data has been collected from BTS, United States Department of Transportation. The data included all the flights operated by American Airlines, connecting the top five busiest airports of United States, located in Atlanta, Los Angeles, Chicago, Dallas/Fort Worth, and New York, in the years 2015 and 2016. Aforesaid supervised machine learning algorithms were evaluated to predict the arrival delay of individual scheduled flights. All the algorithms were used to build the predictive models and compared to each other to accurately find out whether a given flight will be delayed more than 15 min or not. The result is that the gradient boosting classifier gives the best predictive arrival delay performance of 79.7% of total scheduled American Airlines’ flights in comparison to kNN, SVM and random forest. Such a predictive model based on the GBC potentially can save huge losses; the commercial airlines suffer due to arrival delays of their scheduled flights.