Knowledge-Driven Multi-Label Classification Of Process Scheduling Problems

Process scheduling problems have been largely studied in the literature, and a large number of methods and approaches are available and capable of solving them. However, the selection of the best method to a real practical problem at hand and the limited number of experts in optimization and operations research within industrial environments seriously limit the practical application of the theoretical methods. This work proposes a framework, based on a multi-label classification strategy, for selecting those mathematical scheduling models that are more suitable to solve a certain scheduling problem definition. We have decomposed the problem description into binary classification problems, in order to analyze the convenience of each scheduling model for a certain definition. As a result, a systematic approach to scheduling model selection is achieved, facilitating the bridge between theoretical developments and industrial practice.