On Application Of Structural Decomposition For Process Model Abstraction


Published on

Real world business process models may consist of hundreds of elements and have sophisticated structure. Although there are tasks where such models are valuable and appreciated, in general complexity has a negative influence on model comprehension and analysis. Thus, means for managing the complexity of process models are needed. One approach is abstraction of business process models—creation of a process model which preserves the main features of the initial elaborate process model, but leaves out insignificant details. In this paper we study the structural aspects of process model abstraction and introduce an abstraction approach based on process structure trees (PST). The developed approach assures that the abstracted process model preserves the ordering constraints of the initial model. It surpasses pattern-based process model abstraction approaches, allowing to handle graph-structured process models of arbitrary structure. We also provide an evaluation of the proposed approach.

Published in: Technology, Business
1 Like
  • Be the first to comment

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

On Application Of Structural Decomposition For Process Model Abstraction

  1. 1. On Application of Structural Decomposition for Process Model Abstraction Artem Polyvyanyy Sergey Smirnov Mathias Weske BPSC 2009 24 March 2009
  2. 2. Motivation <ul><li>Research project with AOK Brandenburg </li></ul><ul><li>Goal: detailed process models abstract process models </li></ul><ul><li>* example model : </li></ul><ul><ul><li>> 300 nodes </li></ul></ul><ul><ul><li>>150 functions </li></ul></ul><ul><ul><li>graph-structured model </li></ul></ul><ul><li>≈ 4 000 EPCs </li></ul><ul><li>graph-structured process models </li></ul>
  3. 3. Business Process Model Abstraction <ul><li>… is generalization of a model, leaving out insignificant process details in order to reduce model complexity and retain information relevant for a particular purpose. </li></ul>
  4. 4. Business Process Model Abstraction What model elements are insignificant? <ul><li>out of scope </li></ul><ul><li>possible criteria: </li></ul><ul><ul><li>non-functional properties </li></ul></ul><ul><ul><li>semantics </li></ul></ul><ul><li>SESE decomposition of a model </li></ul><ul><li>abstraction mechanism </li></ul>How to abstract insignificant elements?
  5. 5. Process Model <ul><li>(N, E, type) is a business process model, where: </li></ul><ul><ul><li>is a set of nodes, where N A ≠ Ø – a set of activities; N G – a set of gateways; the sets are disjoint </li></ul></ul><ul><ul><li>is a set of directed edges between nodes representing control flow </li></ul></ul><ul><ul><li>is a connected graph </li></ul></ul><ul><ul><li>every activity has at most 1 incoming & at most 1 outgoing edge </li></ul></ul><ul><ul><li>there is at least 1 activity with no incoming edges (start activity) and at least 1 activity with no outgoing edges (end activity) </li></ul></ul><ul><ul><li>assigns control flow construct to a gateway </li></ul></ul><ul><ul><li>every gateway is either a split or a join; splits have exactly 1 incoming edge and at least 2 outgoing; joins have at least 2 incoming edges and exactly 1 outgoing. </li></ul></ul>
  6. 6. Aggregation vs. Elimination Aggregate Eliminate
  7. 7. Assumption: Sound Process Models Hidden deadlock Hidden unsafe process fragment unsound model sound model Assume initial models to be sound
  8. 8. Stepwise Abstraction
  9. 9. Order Preserving Abstraction F A A and B belong to F A , ordering constraints are lost C and D do not belong to F A , ordering constraints are preserved A belong to F A and D does not, ordering constraints between F and D as between A and D
  10. 10. Single Entry Single Exit Fragment <ul><li>SESE fragment is a fragment which has exactly: </li></ul><ul><li>1 incoming edge </li></ul><ul><li>1 outgoing edge </li></ul>
  11. 11. Canonical SESE Fragment <ul><li>canonical SESE fragments </li></ul><ul><li>non-canonical SESE fragments </li></ul>
  12. 12. Relations between SESE Fragments p arent - child predecessor-successor if the node set of SESE fragment f 1 is the subset of node set of SESE fragment f 2 , then f 1 is the child of f 2 and f 2 is the parent of f 1 SESE fragment f 1 precedes SESE fragment f 2 (and f 2 succeeds f 1 ) if the outgoing edge of f 1 is the incoming edge of f 2 P 1 c 2 c 1 p 1 s 2 s 1 p 2
  13. 13. Process Structure Tree parent-child predecessor-successor
  14. 14. Auxiliary Concepts <ul><li>A – an activity to be abstracted </li></ul><ul><li>sese A – canonical SESE fragment containing A (is a leaf in the PST) </li></ul><ul><li>sese min – a minimal canonical SESE fragment containing A and at least one more activity ( sese min ≠ sese A ); there are 2 options for sese min : </li></ul><ul><li>there is canonical sese fragment sese A’ which is in predecessor-successor relation with sese A ; then sese min is a SESE fragment with the incoming edge of the predecessor and the outgoing edge of the successor </li></ul><ul><li>if 1 does not hold, than sese min is a SESE fragment which is the parent of sese A </li></ul>
  15. 15. Abstraction Algorithm <ul><li>define the set of activities to be abstracted (let it be I A ); </li></ul><ul><li>if I A has elements, select one activity from the set (let it be A ); else go to 8 ; </li></ul><ul><li>find sese min for A; </li></ul><ul><li>remove from I A all the activities which belong to sese min ; </li></ul><ul><li>replace sese min with aggregating activity with the incoming edge of sese min and the outgoing edge of sese min ; </li></ul><ul><li>if necessary, add the new aggregating activity to I A ; </li></ul><ul><li>go to 2 ; </li></ul><ul><li>stop. </li></ul>
  16. 16. Abstraction Smoothness smoothness = 2 smoothness = 2 smoothness = 5 … loss of information is essential and desired … abstraction smoothness quantitatively estimates the information loss produced by one abstraction step
  17. 17. Smoothness Evaluation (I) <ul><li>Experiment with process models: </li></ul><ul><li>50 models </li></ul><ul><li>real world process models </li></ul><ul><li>50 <|N| < 205 </li></ul><ul><li>graph-structured models </li></ul>
  18. 18. Smoothness Evaluation (II) „ Optimistic“ algorithm „ Pessimistic“ algorithm
  19. 19. Conclusions <ul><li>We proposed the structural abstraction approach based on PST, which is: </li></ul><ul><li>order preserving </li></ul><ul><li>handles graph-structured models </li></ul><ul><li>We evaluated the approach regarding smoothness </li></ul>
  20. 20. Future Work What model elements are insignificant? <ul><li>semantics of model elements </li></ul><ul><li>more fine-grained decomposition methods </li></ul><ul><li>prototypical implementation </li></ul>How to abstract insignificant elements?