Process Discovery by Synthesizing Activity Proximity and User’s Domain Knowledge

Abstract

Process mining techniques assist users to automatically infer process models from event logs. However, the result of process model driven by traditional process mining technique may conflict with the knowledge of users due to some real conditions, i.e. alternative activity is selected due to equipment breakdown. First, the use of heuristics may detect inconsistencies caused by bad guess. Second, extraction of all possible ordering of events reflects historical observation that sometimes hinders users to obtain an ideal process model since the activity has some event types. Yet, the current process mining approach is not totally compatible with some aspects such as extra logs behavior and soundness of process model when the process model changes according to user requirements. This paper presents a method for synthesizing activity proximity from event logs in the area of process mining. The method derives a bounded graph that covers the extra-behavior of an event log according to user’s domain knowledge. Another important property is that it produces a graph with considering the proximity among activities that still contains the original behavior of the event log based on event types. The methods described in this paper have been implemented in ProM framework and tested on a set of real process examples.