Solving a Job-Shop Scheduling Problem by an Adaptive Algorithm Based on Learning

A learning stage of scheduling tends to produce knowledge about a benchmark of priority dispatching rules which allows a scheduler to improve the solution quality for a set of similar job-shop problems. Once trained on the sample job-shop problems (usually with small sizes), the adaptive algorithm solves a similar job-shop problem (with a moderate size or a large size) better than heuristics used as a benchmark at the learning stage of scheduling. Our adaptive algorithm does not guarantee to perform as an exact algorithm or better than a more sophisticated heuristic algorithm (like e.g. the shifting bottleneck one) which need a large running time. For an adaptive algorithm with a learning stage, the job-shop scheduling problem is modeled via a weighted mixed (disjunctive) graph with the conflict resolution strategy used for finding an appropriate schedule.