Supporting the Optimized Execution of Business Processes through Recommendations

Suppo ing he Op imized Execu ion o Business P ocesses h ough Recommenda ions I ene Ba ba1,Ba ba aWebe 2, and Ca melo Del Valle1 1Depa amen o de Lenguajes y Sis emas In o m´a icos, Uni e si y o Se ille, Spain {i eneb ,ca melo}@us.es 2Depa men o Compu e Science, Uni e si y o Innsb uck, Aus ia [email p o ec ed] Abs ac . In o de o be able o lexibly adjus a company’s business p ocesses (BPs) he e is an inc easing in e es in lexible P ocess-Awa e In o ma ion Sys ems (PAISs). This inc easing lexibili y, howe e , yp- ically implies dec eased use guidance by he PAIS and hus poses ad- di ional challenges o i s use s. This wo k p oposes a ecommenda ion sys em which assis s use s du ing p ocess execu ion o op imize pe o - mance goals o he p ocesses. The ecommenda ion sys em is based on a cons ain -based app oach o planning and scheduling he BP ac i i ies and conside s bo h he con ol- low and he esou ce pe spec i e. Keywo ds: Flexible P ocess-Awa e In o ma ion Sys em, Decla a i e Business P ocesses, Recommenda ions, Resou ce alloca ion, P edic ion. 1 In oduc ion Nowadays, flexible P ocess-Awa e In o ma ion Sys ems (PAISs) a e equi ed o allow companies o apidly adjus hei business p ocesses (BPs) o changes in he en i onmen [10]. The specifica ion o p ocess p ope ies in a decla a i e way is an impo an s ep owa ds he flexible managemen o PAISs [3]. Due o hei flexible na u e, equen ly se e al ways o execu e decla a i e p ocess models exis . Typically, gi en a ce ain pa ial ace ( eflec ing he cu en s a e o he p ocess ins ances), use s can choose om se e al enabled ac i i ies (i.e., ac i i ies whose execu ion does no iola e any cons ain o only lead o empo- a y iola ions [6]) which ac i i y o execu e nex . This selec ion, howe e , can be qui e challenging since pe o mance goals o he p ocess (e.g., minimiza ion o o e all comple ion ime) should be conside ed, and use s o en do no ha e an unde s anding o he o e all p ocess. Mo eo e , op imiza ion o pe o mance goals equi es ha esou ce capaci ies a e conside ed. The e o e, ecommenda- ion suppo is needed du ing BP execu ion, especially o inexpe ienced use s. The need o use assis ance du ing he execu ion o decla a i e BPs has been picked up in p e ious wo k [9,5]. Exis ing p oposals, howe e , only conside he con ol-flow pe spec i e o ob aining ecommenda ions, bu no esou ces. In o de o add ess his gap and o suppo use s o flexible PAISs du ing p ocess execu ion in op imizing pe o mance goals like minimizing he o e all comple ion ime (i.e., ime needed o comple e all p ocess ins ances which we e planned o a ce ain pe iod), we p opose he gene a ion o op imized enac - men plans. Fo his, ac i i ies o be execu ed ha e o be selec ed and o de ed (planning p oblem [4]) conside ing bo h con ol-flow and esou ce cons ain s (scheduling p oblem [2]) imposed by he decla a i e specifica ion. Fo planning and scheduling (P&S) he ac i i ies in a way ha he p ocess goal is op imized, a cons ain -based app oach is p oposed since cons ain p o- g amming [7] supplies a sui able amewo k o modeling and sol ing p oblems in ol ing P&S [8]. Fo his, he decla a i e model is complemen ed wi h in o ma- ion ela ed o es ima es ega ding he numbe o ins ances, ac i i y du a ions, and esou ce a ailabili ies. Recommenda ions on possible nex s eps a e hen gene a ed conside ing he pa ial ace and he op imized plans. Replanning is suppo ed i ac ual aces de ia e om he op imized plans (e.g., because es i- ma es u ned ou o be inaccu a e). This pape is o ganized as ollows: Sec ion 2 includes an o e iew o ou p oposal, Sec ion 3 shows he applica ion o he p oposed app oach o a unning example, and finally, Sec ion 4 includes conclusions and u u e wo k. 2 Me hod o Gene a ing Recommenda ions To op imize he o e all p ocess pe o mance goals, use s o flexible PAISs a e suppo ed du ing BP execu ion h ough ecommenda ions. A ecommenda ion is composed by one o mo e enabled ac i i ies (i.e., ac i i ies which a e allowed o be execu ed gi en a decla a i e p ocess model and a pa ial ace) o be execu ed nex , oge he wi h hei esou ce alloca ions. Ou p oposal is based on applying op imiza ion echniques du ing bo h build and un- ime (c . Fig. 1). Build- ime. The build- ime phase ocusses on he gene a ion o op imized en- ac men plans om decla a i e BP specifica ions by P&S he ac i i ies. (1) C ea e Decla a i e Speci ica ion. In a fi s s ep, a decla a i e specifica- ion co e ing bo h he con ol-flow and he esou ce pe spec i e o he BP o be suppo ed is c ea ed. We use ConDec [6,11], a decla a i e language which p oposes an open se o cons ain s o he high-le el empla es be ween BP ac i i ies (i.e., exis ence, ela ion and nega ion cons ain s). (2) Ex end Decla a i e Speci ica ion. In o de o P&S he BP ac i i ies, he decla a i e specifica ion is ex ended by conside ing he es ima ed alues o : (i) he du a ion o he BP ac i i ies, (ii) he numbe o ins ances execu ed pe planning pe iod, and (iii) esou ce a ailabili ies. (3) Gene a e Op imized Enac men Plans. Op imized enac men plans a e gene a ed by applying AI echniques o P&S he BP ac i i ies, conside ing he ex ended decla a i e specifica ion. In his wo k, CP is selec ed o he gene a ion o he op imized plans since i supplies a sui able amewo k o modeling and sol ing p oblems in ol ing P&S [8] ( o de ails see [1]). The gene a ed plans con ain in o ma ion abou he numbe o imes each BP ac- i i y is execu ed, he s a and he comple ion imes o each ac i i y execu ion, and he esou ce which is used o each ac i i y execu ion. 1. Decla a i e Speci ica ion (con ol- low and esou ce pe spec i es) 2. Decla a i e Speci ica ion + Es ima es 3. Op imized Execu ion Plans BUILD-TIME P ocess Log Use ex ends Business Expe s ob ains Cons ain -based App oach o Planning Model Sea ch Recommenda ion Se ice Pa ial T ace RUN-TIME 3. Op imized Execu ion Plans 3. Op imized Execu ion Plans ecommenda ion eques ecommenda ion e en s <s a (A1),R01> <comp(A1)> <!R12> ... A CB A ED A EB ... ... ... Decla a i e PAIS Execu ion o P ocess Ins ances Cons ain -based App oach o Replanning Model Sea ch upda es/ op imizes Replanning Module E en Log gi e assis ance Planning Module e j g ki b c d a h Fig. 1. O e iew o ou p oposal Since he gene a ion o op imized plans p esen s NP-complexi y, i is no pos- sible o ensu e he op imali y o he gene a ed plans o all cases. The de eloped cons ain -based app oach [1], howe e , allows sol ing he conside ed p oblems in an efficien way. Despi e he NP-complexi y o he conside ed p oblems, a fi s easible solu ion can be swi ly ound by a g eedy algo i hm. Run- ime. The plans gene a ed in build- ime a e hen used o gi ing ecom- menda ions a un- ime. A un- ime, p ocess ins ances a e execu ed by au ho- ized use s (ain Fig. 1). A any poin du ing he execu ion o a p ocess ins ance, he use can selec om he se o enabled ac i i ies wha o do nex . Howe e , o guide he use o op imize he o e all p ocess goals, ecommenda ions a e p o ided by he ecommenda ion se ice (bin Fig. 1), i.e., p oposing he mos sui able ac i i y o execu e nex 1. Fo his, he ecommenda ion se ice consid- e s he cu en pa ial aces o he p ocess ins ances (cin Fig. 1) and he bes a ailable enac men plan (din Fig. 1) mee ing he cons ain s imposed by he decla a i e specifica ion (ein Fig. 1). As execu ion p oceeds, he enac men o he BP and he esou ce a ailabili- ies a e moni o ed ( in Fig. 1). In pa icula , in o ma ion ega ding s a and comple ion imes o he execu ed ac i i ies, oge he wi h he esou ce a ail- abili ies a e s o ed in he e en log (gin Fig. 1). This in o ma ion is analyzed by he Replanning Module (hin Fig. 1) oge he wi h he op imized plans (i in Fig. 1) o check i plan upda es a e equi ed due o unexpec ed e en s. The 1Fo he cu en wo k, he du a ions o he ecommenda ion eques and he esponse ime a e conside ed negligible compa ed o he du a ion o he p ocess ac i i ies. (VWLPDWHV ,QVWDQFHV  5  5  'XUDWLRQ$  'XUDWLRQ%  'XUDWLRQ&  'XUDWLRQ'  'XUDWLRQ(  2SWLPL]HG3ODQV  $   ( $ ( ' ' 5  5  5  % % & & 5HVRXUFH5HT $UHTXLUHV5 %UHTXLUHV5 &UHTXLUHV5 'UHTXLUHV5 (UHTXLUHV5 $ % ( & ' 6XFF 6XFF 5HVS &KDLQ 5HVS 'HFODUDWLYH6SHFLILFDWLRQ  Fig. 2. Build- ime o he Running Example Replanning Module is in cha ge o upda ing he op imized plans (jin Fig. 1) in wo si ua ions: (1) he e a e some de ia ions, i.e., he execu ion ace is no pa o one o he op imized plans (e.g., he use is no always ollowing he ec- ommenda ions) o es ima es a e inco ec ; and (2) he Replanning Module finds a solu ion which is be e han he cu en op imized plans, since his module is con inuously sea ching o a be e plan by conside ing he e en log du ing BP execu ion, p o ided ha he cu en plan is no op imal. I plan upda es a e equi ed, he Replanning Module needs o access he ex ended decla a i e specifica ion (kin Fig. 1) o gene a e new op imized plans. In gene al, despi e he NP-complexi y o he conside ed p oblems, eplanning is less ime consum- ing han ini ial planning, since mos o he in o ma ion abou p e ious gene a ed plans can usually be eused, and CSP a iable alues become known as execu ion p oceeds. 3 A Running Example In his sec ion, ou app oach is used o gi ing ecommenda ions du ing a hy- po he ical execu ion o a unning example. Figu e 3 shows he build- ime phase o he example. The decla a i e specifica ion includes 5 ac i i ies, A, B, C, D and E, and he ollowing ela ions (ConDec empla es [11]) be ween he ac i i- ies (Fig. 2(1)): Exac ly 1(A), i.e., ac i i y A mus be execu ed exac ly once; Succession(A, B), i.e., o execu e ac i i y B, ac i i y A needs o be exe- cu ed be o e, and ac i i y A mus e en ually be ollowed by ac i i y B; Chain Response(B, C), i.e, immedia ely a e he execu ion o B, C mus be exe- cu ed; Response(A, D), i.e, e en ually a e he execu ion o A, D mus be execu ed; and Succession(D, E), i.e., o execu e ac i i y E, ac i i y D needs o be execu ed be o e, and ac i i y D mus e en ually be ollowed by ac i i y E. Fo he conside ed example, esou ces o wo kinds o oles, R0 and R1, a e conside ed. Fo each BP ac i i y (Fig. 2(1)), a ole is defined. In a nex s ep, he decla a i e specifica ion is ex ended wi h es ima es (Fig. 2(2)). Las ly, he cons ain -based app oach is applied o gene a e op imized enac men plans o he specified p oblem (Fig. 2(3)). He eby, label RIj ep esen s he j- h esou ce wi h ole i, and label Ac k ep esen s he k- h execu ion o ac i i y Ac . Figu e 3 shows he beha io o he ecommenda ion se ice when wo hypo- he ical ins ances wi h gi en aces a e execu ed o he decla a i e specifica ion. A he beginning o he execu ion, plan P1(which has al eady been gene a ed         (QDEOHG$FWLYLWLHV 5HFRP PHQGDWLRQV 5  5  5  ,  ,  VWDUW$  5  ! 3DUWLDO7UDFH RQO FRPSOHWHGDFWLYLWLHVGHSLFWHG ,  ,  $   %  $  FRPS$  ! $  !! $  !$  ! ! !  $  % ! $  !&  '  FRPS$  !5  !VWDUW$  5  !VWDUW%  5  ! FRPS%  !VWDUW&  5  !VWDUW'  5  ! 3  $  %  &  !$  ! FRPS&  ! $  %  &  '  !$  !(  %  FRPS'  !VWDUW(  5  !VWDUW'  5  ! (YHQWV 3  3  $  (  '  &  %  $  (  '  &  %  5  XQDYDLODEOH ! 8SGDWH3ODQWR3  'XUDWLRQRI&  LVWHPSRUDO XQLWVODUJHUWKDQH[SHFWHG !3  LVVWLOOVXLWDEOH 8VHUGRHVQRWIROORZWKH UHFRPPHQGDWLRQ ! 8SGDWH3ODQWR3  7LPHOLQH  3  3  3  Fig. 3. Run- ime o he Running Example du ing build- ime) is conside ed o he ecommenda ions. The op imized plan P1has been c ea ed o wo p ocess ins ances. Ini ially, he pa ial ace o bo h ins ances I1and I2is emp y (column Pa ial T ace, whe e comple ed e en s o ac i i y execu ions a e depic ed). Fu he mo e, ac i i ies A,Cand Do bo h ins ances a e enabled ( eflec ed by whi e ba s), whe eas ac i i ies Band Ea e no enabled ( eflec ed by black ba s). Ac i i ies B1and B2a e no enabled since A mus be execu ed be o e execu ing B (Succession(A, B)). Simila ly, ac- i i ies E1and E2a e no enabled since he execu ion o E equi es a p e ious execu ion o D (Succession(D, E)). Conside ing plan P1, s a ing execu ion o ac i i y A1using esou ce R00is sugges ed. The use ollows he ecommen- da ion. Due o Exac ly 1(A),A1is no enabled anymo e. A ime 1, A1is comple ed, hence ac i i y B1becomes enabled, and he pa ial ace o ins ance I1con ains A1. Fu he mo e, an unexpec ed e en occu s (i.e., esou ce R10be- came una ailable), hence plan P1is no longe alid, and he eplanning module gene a es plan P2. A ime 1, based on plan P2, s a ing execu ion o ac i i y A2 using esou ce R00and B1using esou ce R11is sugges ed. The use ollows he ecommenda ion. Due o Exac ly 1(A),A2is no enabled anymo e. A ime 2, A2is comple ed, hence ac i i y B2becomes enabled. A ime 9, B1is com- ple ed, and s a ing execu ion o ac i i y C1using esou ce R00and D1using esou ce R11is sugges ed. The use ollows he ecommenda ion. A ime 14, C1 is comple ed wo ime uni s la e han expec ed. E en wi h he occu ence o his unexpec ed e en , plan P2is s ill alid due o he slack ime be ween ac i i y C1 and ac i i y E1. A ime 15, D1is comple ed, and ac i i y E1becomes enabled. S a ing execu ion o ac i i y E1using esou ce R00and B2using esou ce R11 is sugges ed. The use pa ially ollows he ecommenda ion, so ha , ins ead o execu ing B2she s a s D2. A e his unexpec ed decision, plan P2becomes in alid, and he eplanning module gene a es plan P3.F omnowon, heBP execu ion p oceeds wi hou de ia ions. 4 Conclusion and Fu u e Wo k We p opose a ecommenda ion sys em o gi ing use s assis ance du ing p ocess execu ion in flexible PAISs o op imize pe o mance goals o he p ocesses (i.e., minimiza ion o o e all comple ion ime). The ecommenda ion sys em is based on a cons ain -based app oach, which is used o P&S he ac i i ies such ha he p ocess goal is op imized. In he p oposed app oach, bo h con ol-flow and esou ces a e conside ed. Fu he mo e, he op imized enac men plans a e up- da ed by eplanning echniques when necessa y. As o u u e wo k, i is in ended o ex end he p oposed app oach by conside ing u he objec i e unc ions. Acknowledgmen s. This wo k has been pa ially unded by he Spanish Min- is e io de Ciencia e Inno aci´on (TIN2009-13714) and he Eu opean Regional De elopmen Fund (ERDF/FEDER). Re e ences 1. Ba ba, I., Del Valle, C.: A Cons ain -based App oach o Planning and Scheduling Repea ed Ac i i ies. In: P oc. COPLAS, pp. 55–62 (2011) 2. B ucke , P., Knus , S.: Complex Scheduling (GOR-Publica ions). Sp inge -Ve lag New Yo k, Inc., Secaucus (2006) 3. Fahland, D., Mendling, J., Reije s, H.A., Webe , B., Weidlich, M., Zugal, S.: Decla - a i e e sus Impe a i e P ocess Modeling Languages: The Issue o Main ainabili y. In: Rinde le-Ma, S., Sadiq, S., Leymann, F. (eds.) BPM 2009. LNBIP, ol. 43, pp. 477–488. Sp inge , Heidelbe g (2010) 4. Ghallab, M., Nau, D., T a e so, P.: Au oma ed Planning: Theo y and P ac ice. Mo gan Kau mann, Ams e dam (2004) 5. Haisjackl, C., Webe , B.: Use Assis ance Du ing P ocess Execu ion – An Expe i- men al E alua ion o Recommenda ion S a egies. In: P oc. BPI (2010) 6. Pesic, M., Schonenbe g, M.H., Sido o a, N., an de Aals , W.M.P.: Cons ain - Based Wo k low Models: Change Made Easy. In: Mee sman, R. (ed.) OTM 2007, Pa I. LNCS, ol. 4803, pp. 77–94. Sp inge , Heidelbe g (2007) 7. Rossi, F., an Beek, P., Walsh, T. (eds.): Handbook o Cons ain P og amming. Else ie (2006) 8. Salido, M.A.: In oduc ion o planning, scheduling and cons ain sa is ac ion. Jou nal o In elligen Manu ac u ing 21(1), 1–4 (2010) 9. Schonenbe g, H., Webe , B., an Dongen, B.F., an de Aals , W.M.P.: Suppo ing Flexible P ocesses h ough Recommenda ions Based on His o y. In: Dumas, M., Reiche , M., Shan, M.-C. (eds.) BPM 2008. LNCS, ol. 5240, pp. 51–66. Sp inge , Heidelbe g (2008) 10. an de Aals , W.M.P., Jablonski, S.: Dealing wi h wo k low change: iden i ica ion o issues and solu ions. IJCSE 15(5), 267–276 (2000) 11. an de Aals , W.M.P., Pesic, M.: Speci ying, disco e ing, and moni o ing se ice lows: Making web se ices p ocess-awa e. In: Technical Repo BPM-06-09, BPM- cen e .o g (2006)