Transfer Learning Operators for Process-oriented Cases

Transfer learning aims to reuse knowledge in a context that is different from the context for which it has been acquired. In Case-based Reasoning (CBR), experiential knowledge is recorded in form of cases. It is a challenging task to port cases across the boundaries of the application domain. In this paper, we propose a framework for the transfer of process-oriented cases from a source to a target domain, including workflows contained in the cases. Adaptation methods for workflows in a classical CBR sense are combined with analogical knowledge learned by means of word embeddings. A set of transformation operators is specified that apply transfer knowledge to a workflow. An experiment with the three operators ’analogical substitution’, ’generalisation’ and ’abstraction’ is conducted. 30 workflows in BPMN from the source domain passenger and baggage handling at the airport are transferred into target domain SAP warehouse management. The experimental results are discussed by means of the 3QM framework which provides metrics to assess the quality of business process.