A New Process Discovery Algorithm for Exploratory Data Analysis

Abstract

The domain of process mining created many discovery techniques which can be used to generate a process representation of the data. However, existing techniques come with a flaw for exploratory data analysis (EDA). They tend to produce process models which contain more process behaviour than is observed in the data and do not optimize for understandability. This severely limits their value for EDA, because only patterns which can be observed from the data should be distilled when performing an EDA. We explain why this limitation is important and give a methodology to overcome this. This methodology describes how a discovery algorithm can be developed that is suitable for EDA.