Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Automated Discovery of Structured Process Models: Discover Structured vs Discover and Structure

918 views

Published on

Research paper presentation at the 35th International Conference on Conceptual Modeling (ER'2016), Gifu, Japan, 15 Nov. 2016
Presentation delivered by Raffaele Conforti.
Paper available at: http://goo.gl/5EN3l2

Published in: Education
  • Be the first to comment

Automated Discovery of Structured Process Models: Discover Structured vs Discover and Structure

  1. 1. CRICOS No. 000213Ja university for the worldreal R Automated Discovery of Structured Process Models: Discover Structured vs Discover and Structure Adriano Augusto, Raffaele Conforti, Marlon Dumas, Marcello La Rosa, and Giorgio Bruno
  2. 2. CRICOS No. 000213Ja university for the worldreal R Automated Process Discovery CID Task Time Stamp … 13219 Enter Loan Application 2007-11-09 T 11:20:10 - 13219 Retrieve Applicant Data 2007-11-09 T 11:22:15 - 13220 Enter Loan Application 2007-11-09 T 11:22:40 - 13219 Compute Installments 2007-11-09 T 11:22:45 - 13219 Notify Eligibility 2007-11-09 T 11:23:00 - 13219 Approve Simple Application 2007-11-09 T 11:24:30 - 13220 Compute Installements 2007-11-09 T 11:24:35 - … … … … Enter Loan Application Retrieve Applicant Data Compute Installments Approve Simple Application Approve Complex Application Notify Rejection Notify Eligibility
  3. 3. CRICOS No. 000213Ja university for the worldreal R Process Quality Dimensions Process Discovery
  4. 4. CRICOS No. 000213Ja university for the worldreal R Process Quality Dimensions Process Discovery Fitness
  5. 5. CRICOS No. 000213Ja university for the worldreal R Process Quality Dimensions Process Discovery Fitness Precision
  6. 6. CRICOS No. 000213Ja university for the worldreal R Process Quality Dimensions Process Discovery Fitness Precision Generalization
  7. 7. CRICOS No. 000213Ja university for the worldreal R Process Quality Dimensions Process Discovery Fitness Precision Generalization Complexity
  8. 8. CRICOS No. 000213Ja university for the worldreal R Process Discovery Algorithms: The Two Worlds High-Fitness High-Precision High-Fitness Low-Complexity
  9. 9. CRICOS No. 000213Ja university for the worldreal R Process Discovery Algorithms: The Two Worlds High-Fitness High-Precision Heuristic Miner Fodina Miner High-Fitness Low-Complexity
  10. 10. CRICOS No. 000213Ja university for the worldreal R Process Model discovered with Heuristics Miner
  11. 11. CRICOS No. 000213Ja university for the worldreal R Process Discovery Algorithms: The Two Worlds High-Fitness High-Precision Heuristic Miner Fodina Miner High-Fitness Low-Complexity
  12. 12. CRICOS No. 000213Ja university for the worldreal R Process Discovery Algorithms: The Two Worlds High-Fitness High-Precision Heuristic Miner Fodina Miner High-Fitness Low-Complexity Inductive Miner Evolutionary Tree Miner
  13. 13. CRICOS No. 000213Ja university for the worldreal R Process Model discovered with Inductive Miner • Structured by construction • Based on process tree
  14. 14. CRICOS No. 000213Ja university for the worldreal R Process Discovery Algorithms High-Fitness High-Precision Low-Complexity
  15. 15. CRICOS No. 000213Ja university for the worldreal R Process Discovery Algorithms High-Fitness High-Precision Low-Complexity Structured Miner
  16. 16. CRICOS No. 000213Ja university for the worldreal R Process Model discovered with Structured Miner
  17. 17. CRICOS No. 000213Ja university for the worldreal R Discover and Structure: A two phases approach • Phase One: discover a process model focussing on fitness and precision without constraints on its structure. For example using Heuristic Miner or Fodina Miner. • Phase Two: simplify the discovered process model structuring it at posteriori.
  18. 18. CRICOS No. 000213Ja university for the worldreal R Phase Two: Structuring Discover the RPST of the model Process Fragment: • Trivial (T) – single edge • Polygon (P) – sequence of fragments • Bond (B) – set of fragments sharing two nodes • Rigid (R) – none of the above cases
  19. 19. CRICOS No. 000213Ja university for the worldreal R Phase Two: Structuring Discover the RPST of the model Reject Payment Request Inform Customer Payby Cash Payby Cheque Approve Update Account P1 P1 B1 B1 P3 R1P2 P2 P3 R1
  20. 20. CRICOS No. 000213Ja university for the worldreal R Phase Two: Structuring Discover the RPST of the model Structure sound AND-Homogeneous or Heterogeneous rigids using BPSTruct (Polyvyanyy 2014)
  21. 21. CRICOS No. 000213Ja university for the worldreal R Phase Two: Structuring Discover the RPST of the model Structure sound AND-Homogeneous or Heterogeneous rigids using BPSTruct (Polyvyanyy 2014) Structure XOR-Homogeneous and unsound rigids using Extended Oulsnam
  22. 22. CRICOS No. 000213Ja university for the worldreal R Oulsnam’s Algorithm Extended for BPMN Process Models • Injection
  23. 23. CRICOS No. 000213Ja university for the worldreal R Oulsnam’s Algorithm Extended for BPMN Process Models • Push-Down – Push down-stream the gateway causing the injection – Duplicate everything in between the gateway causing the injection and the gateway down-stream
  24. 24. CRICOS No. 000213Ja university for the worldreal R Oulsnam’s Algorithm Extended for BPMN Process Models • Ejection
  25. 25. CRICOS No. 000213Ja university for the worldreal R Oulsnam’s Algorithm Extended for BPMN Process Models • Pull-Up – Pull up-stream the gateway causing the injection – Duplicate everything in between the gateway causing the injection and the gateway up-stream
  26. 26. CRICOS No. 000213Ja university for the worldreal R Evaluation Setup • Real-Life dataset: IBM (54 models) and SAP (545 models) collections • Synthetic dataset: 20 models • Generated three sets of logs for a total of 619 logs • We retained all logs for which Heuristics Miner produced an unstructured model - 129 logs
  27. 27. CRICOS No. 000213Ja university for the worldreal R Evaluation Setup • Four process discovery algorithms: – Inductive Miner – Evolutionary Tree Miner – Heuristics Miner – Structured Miner (on top of Heuristics Miner) • Four quality dimensions: – Fitness – Precision – Generalization – Complexity
  28. 28. CRICOS No. 000213Ja university for the worldreal R Evaluation Results • Real-life datasets:
  29. 29. CRICOS No. 000213Ja university for the worldreal R Evaluation Results • Real-life datasets:
  30. 30. CRICOS No. 000213Ja university for the worldreal R Heuristics Miner - Real-life Dataset
  31. 31. CRICOS No. 000213Ja university for the worldreal R Inductive Miner - Real-life Dataset
  32. 32. CRICOS No. 000213Ja university for the worldreal R Structured Miner - Real-life Dataset
  33. 33. CRICOS No. 000213Ja university for the worldreal R Future Work • Experiment with alternative discovery algorithms to explore alternative tradeoffs between model quality metrics • Explore the option of sacrificing weak bisimilarity to obtain models with higher structuredness • Use process model clone detection techniques to refactor duplicates introduced by the structuring phase
  34. 34. CRICOS No. 000213Ja university for the worldreal R Questions ?

×