To prevent spam users, you can only post on this forum after registration, which is by invitation. If you want to post on the forum, please send me a mail (h DOT m DOT w DOT verbeek AT tue DOT nl) and I'll send you an invitation in return for an account.

Process trees and non-free choice constructs

bendalmas
edited October 2015 in Research
Hi everyone !
After reading papers about the importance of the representational bias in process mining, process trees are told to be more suitable to model sound model, even taking into account most of the process modeling requirements (OR-constructs, concurrency, etc...). But they doesn't seem to be effective towards non-free choice constructs. 

In domains such as healthcare, where we often face the situation where a patient goes through a certain activity depending on which one(s) he has already executed, doesn't this unability to handle the non-free choice requirement impact the quality of the model built ? (more than a petri net or a bpmn model would have done in the first place with their unsound model, and if we do not consider the duplication of activity in the model as a good practice)

Benjamin Dalmas
PhD Student - Blaise Pascal University - Clermont-Ferrand

Comments

  • Dear Benjamin,

    Interesting and valid question.

    It is my strong opinion that unsoundness of a model is a crucial reduction in the quality of the process model since it makes is practically unusable.
    You are right in stating that process trees, and block structures in general, cannot express long-term dependencies without falling back to including the data dimension (but which brings back unsoundness). However, you also have to wonder which algorithm would be able to discover such a long term dependency.

    I'm not stating process trees (or the ETM for that matter) are the definitive solution to all problems, but they have less severe issues than most other modelling formalisms :)
    Joos Buijs

    Senior Data Scientist and process mining expert at APG (Dutch pension fund executor).
    Previously Assistant Professor in Process Mining at Eindhoven University of Technology
  • Dear Joos,

    I see your point. The difficulty to face here is to discover a way to handle both long term dependency and perfect soundness in the same model (maybe the petri nets unsoundness could be handled with extensions of petri nets such as inhibitor arcs).

    I think the question is how much importance do we give to this or that requirement in the construction of the model. Maybe it will help us decide with tools is more suitable.


    Benjamin Dalmas
    PhD Student - Blaise Pascal University - Clermont-Ferrand
Sign In or Register to comment.