Investigation of hidden markov model for the tuning of metaheuristics in airline scheduling problems

The tuning approach consists in finding the most suitable configuration of an algorithm for solving a given problem. Machine learning methods are usually used to automate this process. They may enable to construct robust autonomous artifacts whose behavior becomes increasingly expert. This paper focuses on the restriction of this general problem to the field of air planning and more specifically the crew scheduling problem. Metaheuristics are widely used to solve this problem. Our approach consists of using hidden markov model to find the best configuration of the algorithm based on the estimation of the most likely state. The experiment consists of finding the best parameter values of the particle swarm optimization algorithm for the crew scheduling problem. Our approach has shown that it can be a promising solution for automatic optimization of airline scheduling problems. © 2016