A knowledge discovery process for a flexible manufacturing system

We develop a learning system to schedule a workshop of production as well as possible. The elaboration of this tool joined the same problems which are posed in the community working on knowledge data discovery (KDD). Initially we introduce the need for making a learning system to schedule a workshop, then we describe the specificity of our system. We develop 3 phases of the process of KDD and we underline our contribution in the first 2 stages: the data warehousing and the filtering methods. We present a new criterion of data selection and put it in perspective among the existing criteria. We elaborate a filtering algorithm which prepares the data at the step of data mining which uses the software C4/5 to extract knowledge in form of a decision tree. Our work is in supervised machine learning and is applied to the case of a flow shop of 4 machines, where the knowledge concerns the choice of the best heuristic in the workshop.