Similarity Determination in Activity Sequences - A Supportive Framework

An increasing number of information systems support their users by helping them in reusing existing knowledge and experience. Often this is done by retrieving similar instances like similar documents, similar process executions or similar persons. While the recommendations use similarity as central concept, the selection of a suitable measure is often done by intuition. This paper introduces a framework that supports the application engineer in selecting and configuring a suitable similarity measure. The requirements of the intended framework are gathered before the architectural implications are detailed. The resulting framework is applied in a case study in which project performance prediction is to be supported by the similarity of the projects’ activity sequences. The results show the framework’s utility by allowing a comparably simple configuration to yield a considerable support in selecting and configuring a suitable similarity measure.