Meronymy-based Aggregation of Activities in Business Process Models

Transcription

1 Meronymy-based Aggregation of Activities in Business Process Models Sergey Smirnov 1, Remco Dijkman 2, Jan Mendling 3, and Mathias Weske 1 1 Hasso Plattner Institute, Germany 2 Eindhoven University of Technology, The Netherlands 3 Humboldt-Universität zu Berlin, Germany

2 Motivation 2 > 300 nodes > 150 activities Meronymy-based Aggregation of of Activities in in Business Process Models

3 Business Process Model Abstraction 3 is an operation on a business process model preserving essential process properties and leaving out insignificant process details in to retain information relevant for a particular purpose

4 BPMA Scenario 4 Abstraction objects = Activities Abstraction operation = Aggregation

5 Structural BPMA Challenges 5 P Candidate 2 Generate summary Review summary Update customer file Candidate 1? Candidate 1 Candidate 2

6 Activity Ontology 6 Create report Send otification Process summary Update customer profile Generate summary Review summary

7 Role of Activity Ontology in BPMA 7 P Candidate 2 Generate summary Review summary Update customer file Candidate 1? Candidate 1 Candidate 2 Send otification Generate summary Create report Process summary Review summary Update customer profile

8 Aggregation Mining Idea 8 Input: Process model + Ontology Output: Aggregations Algorithm Sketch: How to find an aggregation candidate efficiently? FOR each aggregation candidate map each aggregation candidate activity to an ontology activity IF (ontology activities are strongly related) aggregation candidate is an aggregation How to judge on ontology activity relatedness?

9 Activity Alphabet 9 Create report Send otification Process summary Update customer profile Generate summary Review summary

10 Process Model 10 is a process model, where: finite non-empty set of activities finite set of gateways finite set of nodes the flow relation a connected graph

11 Aggregation Candidate 11 In process model is an aggregation candidate.

12 Meronymy Tree 12 Meronymy tree is a tuple n 0

13 Meronymy Forest 13 Meronymy forest F is a disjoint union of meronymy trees n 0 n 1 n 2 g n 3 n 13 n 4 n 5 e f n 6 n 7 n 8 n 14 n 15 n 16 n 9 n 17 n 18 n 19 n 20 n 21 n 10 n 12

14 Aggregation Candidate Construction 14 Construction of aggregation candidates aggregation through aggregation candidate size increment start: k =2 i iteration: construct i-size aggregation candidates stop: from (i-1) aggregation candidates prune insignificant candidates prune candidates with large distance k= A, PM = (A, G, E) OR all the aggregation candidates of size k are pruned

15 Activity Match (1) 15 Process model Meronymy forest n 0 n 1 n 2 g n 3 n 4 n 5 e f n 6 n 7 n 8

16 Activity Match (2) 16 Process model Meronymy forest n 0 n 1 n 2 g n 3 n 4 n 5 e f n 6 n 7 n 8

17 Activity MixMatch 17 Process model Meronymy forest n 1 g n 0 n 1 n 14 n 15 g n 1 n 2 g n 3 n 14 n 15 n 4 n 5 e f n 6 n 7 n 8

18 Lowest Common Ancestor 18 Lowest common ancestor is a function n 0 maps a tree node set to its lowest common ancestor n 2 n 1 n 2 g n 3 n 4 n 5 e f n 6 n 7 n 8

19 Meronymy Leaves 19 Meronymy leaves is a function n 0 maps an activity to the leaves of the subtree rooted to this activity n 2 n 1 n 2 g n 3 n 4 n 5 e f n 6 n 7 n 8

20 Degree of Aggregation Coverage (1) 20 Degree of aggregation coverage is a function n 0 n 1 n 2 g n 3 n 4 n 5 e f n 6 n 7 n 8

21 Degree of Aggregation Coverage (2) 21 Degree of aggregation coverage is a function n 0 n 1 n 2 g n 3 n 4 n 5 e f n 6 n 7 n 8

22 Degree of Aggregation Coverage Properties 22 Shows if the LCA has other descendents, except aggregation candidate Ignores n 1 n 2 g n 3 aggregation candidate size the aggregation candidate n 4 n 5 e f n 6 n 7 n 8 depth in the LCA subtree ignore the LCA depth Possesses value between 0 and 1 n 0

23 Object Studied in Evaluation 23 Model collection 6 process models (42 activities on average) Meronymy forest MIT Process Handbook processes elicited in the interviews with process experts 5000 activities specifies meronymy and hyponymy spans several business domains

24 Evaluation Approach 24 Each activity aggregation is decomposed into a set of subsets of size 2, e.g.: {a, b, c} {a, b}, {a, c}, {b, c} Modeling expert evaluates pair relevance Experiments varying node distance, cover Observe the precision value

25 Evaluation Results 25 Observations precision 0.46 cover precision node distance precision cover threshold 0.3 cover threshold 0.2

26 Conclusion 26 Contributions Metric for relatedness of activity sets Activity aggregation mining algorithm Future work Improve activity matching technique Precise aggregation mining technique evaluation Investigate other information enabling activity aggregation

27 27 Thank you!

28 Contact Details 28 Sergey Smirnov PhD Student Business Process Technology Group Hasso Plattner Institute Prof.-Dr.-Helmert-Str. 2-3, Potsdam, Germany Phone: +49 (0) Fax: +49(0)