Can machines learn how to forecast taxi-out time? A comparison of predictive models applied to the case of Seattle/Tacoma International Airport

This study compares the performance of ensemble machine learning, ordinary least-squared and penalized algorithms to predict taxi-out time at two different periods of NextGen capability implementation. In the pre-sample, ordinary least-squared and ridge models performed better than other ensemble learning models. However, the gradient boosting model provided the lowest root mean squared errors in the post-sample. No algorithm fits data better in all cases. This paper recommends selecting the model that provides the best balance between bias and variance.

Language

  • English

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 01686504
  • Record Type: Publication
  • Files: TRIS
  • Created Date: Oct 17 2018 3:04PM