Mining Constraints for Ar.ul Processes

Transcription

1 Mining Constraints for Ar.ul Processes Claudio Di Ciccio and Massimo Mecella Claudio Di Ciccio 15 th Interna?onal Conference on Business Informa?on Systems (BIS 2012) Tuesday, May the 22 nd, Vilnius, Lithuania

2 Process Mining Definition Process Mining [Aalst2011.book], also referred to as Workflow Mining, is the set of techniques that allow the extrac?on of process descrip?ons, stemming from a set of recorded real execu?ons (logs). ProM [AalstEtAl2009] is one of the most used plug- in based sovware environment for implemen?ng workflow mining (and more) techniques. The new version 6.0 is available for download at

3 Process Mining Definition Process Mining involves: Process discovery Control flow mining, organiza?onal mining, decision mining; Conformance checking Opera?onal support We will focus on the control flow mining Many control flow mining algorithms proposed α [AalstEtAl2004] and α ++ [WenEtAl2007] Fuzzy [GüntherAalst2007] Heuris?c [WeijtersEtAl2001] Gene?c [MedeirosEtAl2007] Two- step [AalstEtAl2010]

4 A different context Artful processes and knowledge workers Artful processes [HillEtAl06] informal processes typically carried out by those people whose work is mental rather than physical (managers, professors, researchers, engineers, etc.) knowledge workers [ACTIVE09] Knowledge workers create artful processes on the fly Though artful processes are frequently repeated, they are not exactly reproducible, even by their originators, nor can they be easily shared.

5 A different context conversations In collaborative contexts, knowledge workers share their information and outcomes with other knowledge workers E.g., a software development mgr. Typically, by means of several conversations conversations are actual traces of running processes that knowledge workers adhere to

6 A different context Processes from conversations From the collection of messages, you can extract the processes that lay behind Related conversations are traces of their runs Valuable advantages for users Automated discovery of formal representations with no effort for knowledge workers Tidy organization for naïve best practices kept only in mind Opportunity to share and compare the knowledge on methodologies Automated discovery of bottlenecks, delays, structural defects from the analysis of previous runs conversations are a kind of semi-structured text this approach is not tailored to the electronic mail it can be extended to the analysis of other semi-structured texts

7 MailOfMine What is MailOfMine? MailOfMine is the approach and the implementa?on of a collec?on of techniques, the aim of which is to is to automa?cally build, on top of a collec?on of messages, a set of workflow models that represent the ar.ul processes laying behind the knowledge workers ac?vi?es. [DiCiccioEtAl11] [DiCiccioMecella/TR12]

8 The MailOfMine approach From the archive to key parts Mail archive Mail Database Conversa?ons Key Parts Mul?- format mail storage plug- in based crawlers [ZardepoEtAl10]- based clustering algorithm [CarvalhoEtAl04] - based filter

9 Key Parts Concatena?on From key parts to processes Tasks [CohenEtAl04, SakuraiEtAl05] - based task extractor [ZardepoEtAl10]- based Ac?vity indicium Processes

10 Tasks MINERful Processes

11 MINERful The mining algorithm in MailOfMine MINERful is a workflow mining algorithm Its input is a collec?on of strings T and an alphabet Σ T Each string t is a trace Each character is an event (enacted task) The collec?on represents the log Its output is a declara8ve process model What is a declara?ve process model?

12 On the visualization of processes The imperative model Represents the whole process at once The most used notation is based on a subclass of Petri Nets (namely, the Workflow Nets)

13 On the visualization of processes The declarative model Rather than using a procedural language for expressing the allowed sequence of ac?vi?es, it is based on the descrip?on of workflows through the usage of constraints the idea is that every task can be performed, except the ones which do not respect such constraints this technique fits with processes that are highly flexible and subject to changes, such as ar.ul processes The notation here is based on [AalstEtAl06, MaggiEtAl11] (DecSerFlow, Declare) If A is performed, B must be perfomed, no maper before or averwards (responded existence) Whenever B is performed, C must be performed averwards and B can not be repeated un?l C is done (alternate response)

14 On the visualization of processes Imperative vs declarative Declara?ve Impera?ve Declarative models work better in presence of a partial specification of the process scheme

15 A real discovered process model Spaghetti process [Aalst2011.book]

16 Declare constraint templates Constraint templates as Regular Expressions (REs)

17 Declare constraint templates Constraint templates as Regular Expressions (REs)

18 MINERful by example A project mee?ng is scheduled We suppose that a final agenda will be commiped ( confirmagenda ) aver that requests for a new proposal ( requestagenda ), proposals themselves ( proposeagenda ) and comments ( commentagenda ) have been circulated. Shortcuts for tasks (process alphabet): p r c Scenario n ( proposeagenda ) ( requestagenda ) ( commentagenda ) ( confirmagenda )

19 MINERful by example Constraints on tasks Existence constraints 1. Participation(n) 2. Uniqueness(n) 3. End(n) Rela?on constraints 4. Response(r,p) 5. RespondedExistence(c,p) 6. Succession(p,n) The agenda 1. must be confirmed, 2. only once: 3. it is the last thing to do. During the compila?on: 4. the proposal follows a request; 5. if a comment circulates, there has been / will be a proposal; 6. aver the proposal, there will be a confirma?on, and there can be no confirma?on without a preceding proposal.

20 MINERful by example Testing by replay In order to validate the algorithm We translate constraints into REs The overall process is expressed by the intersec?on of REs We use a RE- driven random string builder [Xeger] for crea?ng a test- and- valida?on set We analyze the result and evaluate the performances In order to see how it works now We follow a run of MINERful over a string built by Xeger: r r p c r p c r c p c n

21 MINERful by example p p occurred 3?mes in 1 string γ p (3) = 1 n Building the ownplay of p and n For each m 3 γ p (m) = 0 p did not occur as the first nor as the last character g i (p) = 0 g l (p) = 0 γ n (1) = 1 For each m 1, γ n (m) = 0 n occurred as the last character in 1 string g i (n) = 0 g l (n) = 1 r r p c r p c r c p c n r r p c r p c r c p c n

22 MINERful by example Building the interplay of p and n With respect to the occurrence of p, n occurred i. Never before: 3?mes ii. iii. iv. δ p,n (- ) = 3 2 char s aver: 1?me δ p,n (2) = 1 6 char s aver: 1?me δ p,n (6) = 1 9 char s aver: 1?me δ p,n (9) = 1 v. Alterna?ng: i. Onwards: 2?mes ii. b p,n = 2 Backwards: never b p,n = 0 Looking at the string i. r r p c r p c r c p c n ii. r r p c r p c r c p c n iii. r r p c r p c r c p c n iv. r r p c r p c r c p c n v. i. r r p c r p c r c p c n ii. r r p c r p c r c p c n

23 MINERful by example Building the interplay of r and p δ r,p b r,p = 1 b r,p = 0 r r p c r p c r c p c n

24 MINERful by example Workflow discovery by constraints inference Interplay and ownplay cons?tute the Knowledge Base of MINERful The KB construc?on is such that each new string adds informa8on The algorithm does not need to read the strings more than once each Constraints are determined by the evalua?on of boolean queries on the KB This allows the discovery of constraints with a faster procedure on a smaller set than the whole input MINERful is a two- step algorithm

25 MINERful by example RespondedExistence(c,p) þ (δ r,p (0) > 0) There is no string where p does not occur, if r is read Response(r,p) þ RespondedExistence(r,p) (δ r,p (+ ) > 0) RespondedExistence(r,p) holds and there is no string where p does not follow r Precedence(r,p) ý RespondedExistence(p,r) (δ r,p (- ) > 0) RespondedExistence(p,r) holds and there is no string where p does not precede r Succession(p,n) þ Some queries for inferring constraints Response(p,n) Precedence(p,n)

26 MINERful by example Other inferred constraints Response(c, n) RespondedExistence(c, p) NotSuccession(n, c), NotSuccession(n, p), NotSuccession(n, r) Par?cipa?on(p) AlternatePrecedence(p, n) Succession(p, n) AlternatePrecedence(r, c) Response(r, p)

27 Rela?on constraint templates subsump?on Constraint templates are not independent of each other E.g., A trace like a b a b c a b c c sa?sfies (w.r.t. b and a): RespondedExistence(a, b), RespondedExistence(b, a), CoExistence(a, b), CoExistence(b, a), Response(a, b), AlternateResponse(a, b), ChainResponse(a, b), Precedence(a, b), AlternatePrecedence(a, b), ChainPrecedence(a, b), Succession(a, b), AlternateSuccession(a, b), ChainSuccession(a, b) The mining algorithm would show the most strict constraint only (ChainSuccession(a, b)) MINERful faces and solves this issue, by refining queries on the basis of the subsump?on hierarchy of constraints Mining Constraints for Artful Processes (Di Ciccio, C. - DIAG, SAPIENZA Università di Roma)

28 Rela?on constraint templates subsump?on Constraint templates are not independent of each other Mining Constraints for Artful Processes (Di Ciccio, C. - DIAG, SAPIENZA Università di Roma)

29 Conclusions Recap MailOfMine is a system designed for mining ar.ul processes out of collec?ons MINERful is the worflow mining algorithm designed for MailOfMine MINERful is Independent on the formalism used for expressing constraints Modular (two- phase) Capable of elimina?ng redundancy in the process model

30 Conclusions On the asymptotic complexity of MINERful Linear w.r.t. the number of strings in the testbed T Quadra?c w.r.t. the size of strings in the testbed t max Quadra?c w.r.t. the size of the alphabet Σ T Hence, polynomial in the size of the input O( T t max 2 Σ T 2 )

31 References Cited articles and resources, in order of appearance [Aalst2011.book] van der Aalst, W.M.P.: Process Mining: Discovery, Conformance and Enhancement of Business Processes. Springer (2011). [AalstEtAl2009] van der Aalst, W.M.P., van Dongen, B.F., Güther, C.W., Rozinat, A., Verbeek, E., Weijters, T.: Prom: The process mining toolkit. In de Medeiros, A.K.A., Weber, B., eds.: BPM (Demos). Volume 489 of CEUR Workshop Proceedings., CEUR-WS.org (2009) [AalstEtAl2004] van der Aalst, W.M.P., Weijters, T., Maruster, L.: Workflow mining: Discovering process models from event logs. IEEE Trans. Knowl. Data Eng. 16(9) (2004) [WenEtAl2007] Wen, L., van der Aalst, W.M.P., Wang, J., Sun, J.: Mining process models with non-free-choice constructs. Data Min. Knowl. Discov. 15(2) (2007) [GüntherEtAl2007] Günther, C.W., van der Aalst, W.M.P.: Fuzzy Mining - Adaptive Process Simplification Based on Multi-perspective Metrics. BPM 2007: [WeijtersEtAl2001] Weijters, A., van der Aalst, W.: Rediscovering workflow models from eventbased data using little thumb. Integrated Computer-Aided Engineering 10 (2001) [MedeirosEtAl2007] Medeiros, A.K., Weijters, A.J., Aalst, W.M.: Genetic process mining: an experi- mental evaluation. Data Min. Knowl. Discov. 14(2) (2007) [AalstEtAl2010] van der Aalst, W., Rubin, V., Verbeek, H., van Dongen, B., Kindler, E., Gnther, C.: Process mining: a two-step approach to balance between underfitting and overfitting. Software and Systems Modeling 9 (2010) /s z.

32 References Cited articles and resources, in order of appearance [HillEtAl06] Hill, C., Yates, R., Jones, C., Kogan, S.L.: Beyond predictable workflows: Enhancing productivity in artful business processes. IBM Systems Journal 45(4), (2006) [ACTIVE09] Warren, P., Kings, N., et al.: Improving knowledge worker productivity - the active integrated approach. BT Technology Journal 26(2), (2009) [DiCiccioEtAl11] Di Ciccio, C., Mecella, M., Catarci, T.: Representing and Visualizing Mined Artful Processes in MailOfMine. USAB 2011:83-94 [DiCiccioMecella12] Di Ciccio, C., Mecella,M.: Mining constraints for artful processes. In W. Abramowicz, D. Kriksciuniene, V.S., ed.: 15th International Conference on Business Information Systems. Volume 117 of Lecture Notes in Business Information Processing., Springer (2012) (to appear). [DiCiccioMecella/TR12] Di Ciccio, C., Mecella, M.: MINERful, a mining algorithm for declarative process constraints in MailOfMine. Technical report, Dipartimento di Ingegneria Infor- matica, Automatica e Gestionale Antonio Ruberti SAPIENZA, Universita` di Roma (2012). [AalstEtAl06] van der Aalst, W.M.P., Pesic, M.: Decserflow: Towards a truly declarative service flow language. Proc. WS-FM 2006 [MaggiEtAl11] Maggi, F.M., Mooij, A.J., van der Aalst, W.M.P.: User-guided discovery of declarative process models. In: CIDM, IEEE (2011) [Xeger]