Discovering Process Models from Unlabelled Event Logs

摘要: Existing process mining techniques are able to discover models from event logs where each is known have been produced by a given instance. In this paper we remove restriction and address the problem of discovering model when log provided as an unlabelled stream events. Using probabilistic approach, it possible estimate means iterative Expectaction---Maximization procedure. The same procedure can be used find case id in logs. A series experiments show how proposed technique performs under varying conditions presence certain workflow patterns. Results presented for running example based on technical support process.