Domain-driven actionable process model discovery

Process discovery is a type of process mining that constructs a process model from the event logs of an information system. The model discovered using process discovery techniques and the process as perceived by users will always differ in some ways and to some extents. In particular, less structured process, such as operational process in business and manufacturing, often result overly confusing, spaghetti-like, process models caused by the inherent complexity of the process. As a result, the mined model has many limitations for providing the users with explicit knowledge that can be directly used to influence behavior for the user's interest. Explicit knowledge, as later called by actionable knowledge, is an important representation on measuring the interestingness of mined patterns. This actionable knowledge, which is incorporated with users' background knowledge and based on some notions of actionable rules, can result an actionable process model. Undoubtedly, domain experts, who know the process well, play a key role to enhance the mined model into an actionable model by their involvements during the discovery process. This paper presents a discovery method to obtain an actionable process model that is based on both the event relation in the log and users' knowledge to improve the incompatibility of the traditional process mining approaches. Users can set their knowledge in terms of constraints. Unlike the existing approach, the proposed approach synthesizes the activity proximity and attempts to extract behavior satisfied by the constraints which may be hidden in the event logs for resulting an actionable process model. In addition, the proposed method is used in order to achieve a sound process model when the existence of the constraints does not satisfy the workflow soundness property. The method was implemented in the ProM framework and tested on a real process.