Processes Are Concepts, Aren’t They?

Abstract

Discovery is an information / technical approach to the important managerial problem of decision making under not only uncertainty, but also actually, “unknown unknowns”. Formal Concept Analysis (FCA) is an elegant mathematically grounded theory that complements Data Discovery particularly well, especially for data with no predefined structure. It allows for discovering emergent, a-priori unknown concepts in these data. Process Discovery deals with emergent behaviour in business workflows. In its current state-of-the-art, Process Discovery combines machine-learning techniques utilizing Hidden Markov Model (HMM) representations of Processes. In one of our research lines, we investigate how FCA can improve and complement Process Discovery. Although the inclusion of temporal data and events in FCA allows for the creation of ”early warning” and trend-models, HMM’s are needed for a deep understanding of the processes and their intrinsic complexities. However, FCA can assist significantly in the post-processing and understanding of HMM’s, in particular in the presence of process variations and exceptions. But FCA allows also for the detection of recurrent, coherent Process steps, which can be considered ”service steps” in business processes. Ultimately, an appropriate mathematical representation of HMM’s should allow for the application of algebraic extensions of FCA for discovering Processes and their variations as mathematical concepts as well. Some initial work on Process patterns gives inspiring research directions. Real-life case materials from Healthcare Administration, Customer Contact-Center’s and Financial Services illustrate this keynote lecture.