Several families of core problems in transportation and logistics such as vehicle routing, facility location, and crew scheduling remain formidably challenging to solve for the operations research community and, for most of them, efficient algorithms are still sought after by the industry. One recent research trend explores the possibility of combining optimization and machine learning in innovative ways to provide more accurate models and design improved algorithms. Machine learning and optimization can be applied sequentially or in an integrated fashion. In the former case, machine learning can be used, for example, to estimate some problem input for the optimization model, to preprocess data with the goal of reducing the size of the model to solve, or to describe customer behavior and preferences. In the latter case, machine learning can be applied, for example, to adjust the values of some of the parameters controlling the optimization algorithm or to make heuristic decisions within the algorithm to increase its efficiency.
ARCHETTI, C., CORDEAU, J.F. et DESAULNIERS, G. (2020). Introduction to the special issue on combining optimization and machine learning: Application in vehicle routing, network design and crew scheduling. EURO Journal on Transportation and Logistics, 9(4), pp. 100024.