Deep Learning Driven Optimization for Large Scale Airline Operations

Crew expenses are one of the major cost components of airlines. Integer programming methods were utilized for decades to find solutions that offer a good balance between cost and operational efficiency. However, solving crew pairing problems is still computationally challenging for large-scale schedules. In this project we utilize deep learning based methods that leverage previous optimization results to predict solutions to new problems, increasing the solution speed and quality for a wide variety of airline crew pairing problems.

N. Kemal Ure
ITU Vision Lab
Funded by
April 2018