Foundations of Process Enhancement

Abstract

Abstract Process models are among the milestones for Business Process Management and Mining, and used to describe a business process or to prescribe how its instances should be carried out. It follows that they need to fulfill certain properties to be useful. If they aim to represent how the process is currently being executed, they need to be precise and recall the behavior observed in reality. If the goal is to ensure that the process is executed according to laws and regulations, its model should only allow the behavior that is valid from a domain viewpoint and provides some guarantee to ensure good performance level. Process enhancement is the type of Process Mining that aims at models that fulfill these properties, and the literature further splits it into two subfields: process extension and process improvement. Process extension aims to incorporate the process perspectives on data, decision, resources and time into the model: their inclusion in process models enable designers to fine-tune the model specifications, thus obtaining models with higher levels of precision. Process improvement passes through an “improved” process model. If the model contains portions of behavior that lead to unsatisfactory outcomes (high costs, low customer satisfactions, etc.) or that violate norms and regulations, one would like those portions to be disallowed by the model. In case some executions are observed in reality and are not allowed by the model, they should be incorporated into the model if they are observed to generally yield good performances. This chapter discusses these two types of process enhancement, and illustrates some basic and some advanced techniques to tackle it, highlighting the pros and cons, and the underlaying assumptions.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 25 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.