Decentralized scheduling with dispatching rules is applied in many fields of production and logistics, especially in highly complex manufacturing systems, e.g. semiconductor manufacturing. Nevertheless, no dispatching rule outperforms other rules across various objectives, scenarios and system conditions. In this paper we present an approach to dynamically select the most suitable rule for the current system conditions in real time. We calculate Gaussian process (GP) regression models to estimate each rule’s performance and select the most promising one. The data needed to create these models is gained by a few preliminary simulation runs of the selected job shop scenario from the literature. The approach to use global information to create the Gaussian process models leads to better local decision at the machine level. Using a dynamic job shop scenario we demonstrate, that our approach is capable of significantly reducing the mean tardiness of jobs.