An Industrial Learning Classifier System (LCS) was developed a decade ago for the mining of information in process industries, specifically for a Steel Hot Strip Mill. Despite encouraging results, the prediction accuracy achieved and the effort required did not warrant adoption. The lessons learnt for applying Genetic-based Machine Learning to industrial data-mining applications are still relevant and are described here. After 10 years further research into LCS much innovation has occurred: messy encodings, rule-base reduction, pass-through rules and flexible encodings. However, this paper hypothesises that the biggest hurdle preventing LCS from wider industrial adoption is a lack of ‘abstraction’, i.e., after states have been linked to actions and generalised by removing irrelevant information, similarities between rules must be abstracted to form higher level rules. Initial results for a ‘toy’ problem demonstrate that LCS are capable of abstraction with powerful consequences.
Weitere Kapitel dieses Buchs durch Wischen aufrufen
Bitte loggen Sie sich ein, um Zugang zu diesem Inhalt zu erhalten
Sie möchten Zugang zu diesem Inhalt erhalten? Dann informieren Sie sich jetzt über unsere Produkte:
- Improving Evolutionary Computation Based Data-Mining for the Process Industry: The Importance of Abstraction
William N. L. Browne
- Springer Berlin Heidelberg
in-adhesives, MKVS, Hellmich GmbH/© Hellmich GmbH, Zühlke/© Zühlke