Mining for User-Defined Categorizations as an Approach for Process Simplification in Business Process Discovery

Abstract

Business process discovery approaches analyze event logs to create process models describing the current state of the underlying processes. Because this type of process mining is especially relevant for low-structured processes, there are approaches designed to deal with such processes by simplifying the resulting model. Such simplifications are primarily applied with metrics based on the frequency of observed behaviours. However, a high frequency of a certain behaviour is not synonymous with high relevance to the user. Consequently, this paper applies a design science research approach to develop and implement a business process discovery approach mining for user-defined categories to guarantee relevance to the respective user. This categorization approach is demonstrated in a case study that serves as the basis for an evaluation regarding its perceived usefulness in an expert survey. Although the method currently requires the user to categorize activities before simplifying, it is perceived to be useful by experts.