Study of SVM-based Intelligent Dispatcher for Parallel Machines Scheduling with Sequence-dependent Setup Times

Because the parameter tuning of dispatching rule (DR) is a tedious work for human experts, we have developed a support vector machine (SVM) based intelligent dispatcher to schedule jobs at each decision point for parallel machines scheduling with sequence-dependent setup times. The optimization target is to minimize the total weighted tardiness(TWT). The grid search method is used to find the most appropriate values of the scaling parameters (k1, k2) for the apparent tardiness cost with setups (ATCS) rule to produce good solutions. Besides, the dispatching trajectory of the solution with the most appropriate combinations of k1 and k2 is recorded as the training set. Since SVM provides good generalization for high dimensional input data or few training examples [1], we employ this machine learning approach to extract patterns from the characteristics of a given dispatching trajectory. SVM is then employed to develop a intelligent task comparator. Comparison of simulation results show that the proposed approach has a good generalization performance for unobserved scenarios.