Learning dispatching rules via an association rule mining approach

This thesis proposes a new idea using association rule mining-based approach for discovering dispatching rules in production data. Decision trees have previously been used for the same purpose of finding dispatching rules. However, the nature of the decision tree as a classification method may cause incomplete discovery of dispatching rules, which can be complemented by association rule mining approach. Thus, the hidden dispatching rules can be detected in the use of association rule mining method. Numerical examples of scheduling problems are presented to illustrate all of our results. In those examples, the schedule data of single machine system is analyzed by decision tree and association rule mining, and findings of two learning methods are compared as well. Furthermore, association rule mining technique is applied to generate dispatching principles in a 6 x 6 job shop scheduling problem. This means our idea can be applicable to not only single machine systems, but also other ranges of scheduling problems with multiple machines. The insight gained provides the knowledge that can be used to make a scheduling decision in the future.