Семинар ПОИС: "Asynchronous Interaction Patterns for Mining Process Models of Multi-Agent Systems from Event Logs".

This work addresses the problem of discovering structured and sound process models from event logs produced by multi-agent systems. We focus on analyzing asynchronous agent interactions formally expressed in terms of channel-composed workflow nets. By using morphisms, we prove that the channel-composition preserves soundness of component nets. We implement this formal compositional approach with the help of service interaction patterns. They describe typical ways of asynchronous communication. In our study, two-agent interaction patterns are considered. We also show how these patterns are applied within process discovery. The experimental evaluation shows the increase in the quality of models discovered compositionally via interaction patterns.