Efficient event handling in supply networks using Q-learning and K-means clustering

Abstract

Modern value-added processes will be globally cross-linked through outsourcing and reduction of real net output ratio. Therefore logistical planning and control processes become more complex. Events in supply networks and their consequences to the partners in the supply network will be hardly to overlook without using computer based decision support systems. This paper describes such a decision support system, learning the rules used to control the production network. In details the system architecture will be described, requirements to such a system will be identified and a solution developed at the Heinz Nixdorf Institute and Fraunhofer ALB (application center for logistic-oriented business administration) in Paderborn will be presented. The solution is based on a q-learning approach supported by a k-means clustering algorithm. © 2008 International Federation for Information Processing.