Applying Machine Learning for Solver Selection in Scheduling: A Case Study

We investigate the automated algorithm selection for a workforce scheduling problem that is solved by two different approaches. The solver based on constraint programming techniques has several advantages and it has been used successfully in the industry. However, this algorithm can not solve very large instances in a reasonable amount of time. The metaheuristic solver overcomes this limitation and is able to find solutions even for huge real world instances. We apply machine learning algorithms to select the best suited solver for a particular instance based on problem features. The preliminary experimental results on application of different learning techniques are presented.