Data Transformation and Semantic Log Purging for Process Mining

Existing process mining approaches are able to tolerate a certain degree of noise in process log. However, processes that contain infrequent paths, multiple (nested) parallel branches, or have been changed in an ad-hoc manner, still pose challenges. For such cases, process mining typically returns “spaghetti-models”, that are hardly usable even as a starting point for process (re-)design. In this paper, we address these challenges by introducing data transformation and pre-processing steps that improve and ensure the quality of mined models for existing process mining approaches. We propose the concept of semantic log purging, i.e., the cleaning of logs based on domain specific constraints utilizing knowledge that typically complements processes. Furthermore we demonstrate the feasibility and effectiveness of the approach based on a case study in the higher education domain. We think that semantic log purging will enable process mining to yield better results, thus giving process (re-)designers a valuable tool.