Decision-Making Support for the Correctness of Input Data at Runtime in Business Processes

DECISION-MAKING SUPPORT FOR THE CORRECTNESS OF INPUT DATA AT RUNTIME IN BUSINESS PROCESSES MAR´IA TERESA G ´OMEZ-L ´OPEZ and RAFAEL M. GASCA ∗ Depa men o de Lenguajes y Sis emas In o m icos, Uni e si y o Se ille Se ille, Spain JOSE ´ MIGUEL P´EREZ- ´ALVAREZ † In ellimen Secu i y Toma es, Se ille, Spain In a business p ocess, he in o ma ion ha lows be ween he ac i i ies can be in oduced by hose use s who in e ac wi h he p ocess. This in oduced in o ma ion could be inco ec due o a lack o knowledge o a mis ake. Fo his eason and o make he business p ocess execu ion consis en , we p opose a Decision Suppo Sys em (DSS) o in o m he use abou he possible and co ec alues ha he inpu da a can ake. The DSS akes in o accoun he business p ocess model and he policy o he company. The policy conce ning he inpu da a and da a low ha he company manages can be ep esen ed by cons ain s (called Business Da a Cons ain s). In o de o asce ain all he possible alues o he inpu da a ha pe mi he execu ion o he p ocess ollowing he de ined goals, he DSS analyses he business p ocess model and he Business Da a Cons ain s, using he cons ain p og amming pa adigm. Keywo ds: Business p ocesses; Inpu Da a; Decision-making suppo ; Nume ical ech-niques; P ocess Ins ance Con o mi y. 1. In oduc ion A business p ocess consis s o a se o ac i i ies ha a e pe o med in coo dina- ion wi hin an o ganiza ional and echnical en i onmen 1. In a business p ocess, he in o ma ion ha lows be ween he ac i i ies can be in oduced by he use s. Some imes, he use mus decide he alue o in oduce while aking in o accoun he po en ial ac ions in o de o make he p ocess ins ance co ec . In he business p ocess scena io, his implies he analysis o all he possible b anches ha can be execu ed, and he decisions ha can be aken in he u u e. I he decision made is inco ec , i will a ec o he decisions in he u u e, o i may e en make i impos- ∗{may egomez, gasca}@us.es. www.lsi.us.es/∼qui i †jmpe ez@in ellimen sec.com, h p://www.in ellimen sec.com/ 1 sible o inish he ins ance co ec ly ( ollowing he goals de ined o he company). The equi emen o decision suppo equen ly a ises when decisions ha e o be made in complex, unce ain, and/o dynamic en i onmen s 36. Fo his eason, we p opose a solu ion whe e he decision-making suppo o inpu da a can be in e- g a ed in o he business p ocess ins ances o in o m he use abou he possible alues o he inpu a iables, he eby making he ins ance o he business p ocess consis en . F om he poin o iew o inpu da a alues, he co ec ness o a busi- ness p ocess is based on he co ec ness o he compliance ules ha desc ibe he policy o he company; a co ec inpu alue is he e o e he alue o a a iable ha sa is ies all he ules de ined by he company. Nume ous s udies p opose a a ie y o axonomies o classi y he de ini ion o business compliance ules 2, 3, 4, 5, such as a speci ica ion, a policy o a s an- da dized p ocedu e, ha ep esen a na u al s ep owa ds he inclusion o seman ic equi emen s be ween business unc ionali y and da a. Howe e i he ela ion be- ween he alues o he da a low a iables has o be desc ibed, hen Business Da a Cons ain s (hence o h e e ed o as BDCs) ha e he capaci y o desc ibe Business Compliance Rules 6. These BDCs a e unde s ood as a ype o business compliance ules, which ep esen he seman ic ela ion be ween he da a alues ha a e in o- duced, ead and modi ied du ing he execu ion o he business p ocess ins ances 7. The use o BDCs in decision-making suppo can be decisi e, since humans mus o en make decisions abou he inpu da a in a business p ocess ins ance ha may esul in being inco ec o he wo king o de o he p ocess. I all he po en ial scena ios in he u u e a e aken in o accoun in he decision-making suppo o inpu da a, hen la e iden i ica ion o non-con o mi ies o inpu da a wi h espec o he BDCs can be p e en ed. The e o e, we p opose he use o BDCs o assis in he decisions o he use s, despi e he complica ion ha he BDCs can be associa ed wi h one ac i i y, a se o ac i i ies, o o he whole business p ocess, and hence he assis ance mus in eg a e he model o he business p ocess. The BDCs espond o he demand o he p ocessing o da a by p o iding mo e seman ic con en in he business p ocess, which leads o a be e unde s anding o he p ope ies o he da a used in he business p ocess. In o de o explain ou p oposal, an example o web se ices selec ion is de- pic ed in Figu e 1, whe e a e y simple business p ocess model is shown. We ha e used he BPMN 2.0 no a ion 13 o inpu and ou pu da a. The BDCs a e ep- esen ed by means o anno a ions in BPMN associa ed o he ac i i ies. In his p ocess, he alues o da a i ems DE (Da a Enc yp ion) and NR (max Numbe o Reques s) a e decided and in oduced by a human in o di e en ac i i ies, since hey a e a iables ha can pa icipa e in a decision-making p ocess. The alue o TS (Th oughpu Sa u a ion) is de i ed by Selec Se ice Le el, and hence canno be de ined as an objec i e o he decision-making p ocess. Depending on he alue o DE (DE≤2 o DE>2), one o he b anches o he p ocess is execu ed (XOR ga eway o he di e en p o ide s o se ices). An example o inpu da a decision could in- Selec Se ice Le el Selec P o ide A Selec P o ide B DE<=2 DE>2 Da a Enc yp ion (DE) Th oughpu Sa u a ion (TS) Max Numbe o Reques s (NR) Op ion 1 Op ion 2 0<=DE<=3 300<=TS<=2000 TS<=400*DE DE:1..3 (0<=DE<=3 300<=TS<=2000 TS<=400*DE) AND (DE>=2  NR<=TS 500<=NR<=2000) OR (NR<=1.3*TS 600<=NR<=2000) DE:2..3 NR<=TS 500<=NR<=2000 NR<=1.3*TS 600<=NR<=2000 0<=DE<=3 300<=TS<=2000 TS<=400*DE Fig. 1. Simple example o Business P ocess Ins ance o decision-making suppo o inpu da a suppo ol e he de e mina ion o he possible co ec alues o DE in Selec Se ice Le el ac i i y. As men ioned abo e, each BDC can be associa ed wi h one ac i i y, a se o ac i i ies, o o he whole p ocess: he e o e each ac i i y has se e al BDCs asso- cia ed, o example Selec Se ice Le el is associa ed o {0≤DE≤3, 300≤TS≤2000, TS≤400*DE}. In o de o decide he co ec alues ha he a iable DE can ake o sa is y all he BDCs in he u u e, ei he hose BDCs ela ed wi h only Selec Se ice Le el (Op ion 1 o Figu e 1) can be aken in o accoun , o all he BDCs ela ed di ec ly o indi ec ly wi h DE. This implies analysing all he possible pa hs om Selec Se ice Le el, while conside ing he condi ions associa ed wi h he con- ol lows and e alua ed a un ime (DE in he example), and he BDCs o Selec P o ide A and Selec P o ide B (Op ion 2 o Figu e 1). Op ion 1 p esen s an in e al wi h he possible alues o he a iable DE ([1..3]) when solely he BDCs ela ed o he ac i i y Selec P o ide A a e analysed; and Op ion 2 shows how he in e al wi h possible co ec alues o DE a e educed ([2..3]) i he BDCs ela ed o he a iable NR a e also included in he analysis. Fo example, in Op ion 1 i is possible o in oduce he alue 1 o he a iable DE, and o assign 350 o he a i- able NR ({TS ≤400 ∗DE}). The p oblem a ises when he ac i i y Selec P o ide Ais execu ed, and i is disco e ed ha he e a e no alid alues o he a iable NR o sa is y he BDC {NR ≤TS}. F om he p e ious example, i can be obse ed ha he BDCs could help o in oduce consis en da a in business p ocess ins ances. These BDCs cons i u e he basic knowledge o a Decision Suppo Sys em (DSS). This knowledge can be en iched by including he analysis o he business p ocess model, and he condi ions associa ed wi h he sequence lows. This decision-making suppo o inpu da a can also be used o gua an ee he exis ence o a co ec ins ance o he p ocess, he eby ob aining a mo e aul - ole an p ocess. In o de o suppo his assis ance wi h he DSS o inpu da a, he ollowing aspec s ha e been conside ed: •The use o Business Da a Cons ain s o help in he decision- making suppo o inpu da a. BDCs can assis in he decision-making suppo by epo ing on he possible co ec alues o he da a in oduced in each ins ance; howe e no all o he alues o he BDC a iables a e ye known and ins an ia ed. •The de elopmen o an algo i hm o a e se he p ocess model. To b ing he ele an pa s o he business p ocess oge he and allow hem o con ibu e owa ds decision-making suppo , an analysis o he p ocess model and he BDCs o each ac i i y is necessa y. We p opose an algo i hm ha a e ses he business p ocess model and combines he BDCs ela ed o each ac i i y in o de o ob ain a ep esen a ion o he co ec alues ha he a iables can assume. •The gene a ion o he Nume ical ep esen a ion o he possible co ec alues o he inpu da a. To gene a e a se o quali ied solu ion al e na i es, ins ead o p o iding only one solu ion, we p opose ob aining he possible anges o he decision a iables by means o Cons ain P o- g amming. •The implemen a ion o an applica ion o in eg a e ou p oposal in o business p ocess modelle so wa e. In o de o acili a e he decision-making p ocess desc ip ion, we ha e de eloped an applica ion o he DSS (called MARTIN: MAking Reasoning o daTa INpu ) ha allows he use o de ine he BDCs o each ac i i y, and connec he symbolic and nume ical sol e s wi h he p ocess in o de o ob ain he possible co ec alues o he inpu da a in each ins ance. This applica ion o e s he e- qui ed agili y and lexibili y o he o ganiza ions so ha hey can comply wi h changes in policy and legisla ion, and apply hem in he DSS. Fo hese easons, his pape is o ganized as ollows: Sec ion 2 explains he g am- ma o BDC used in his pape . Sec ion 3 p esen s a mo i a ing example whe e decision-making suppo o inpu da a is used. Sec ion 4 s a es he necessa y de - ini ions o o malize he p oposal. Sec ion 5 analyses how o a e se a business p ocess model o s udy all he possible co ec alues o a a iable, and o p opose an algo i hm o i s de elopmen . Once he BDCs in ol ed in he decision-making suppo a e known, how Cons ain P og amming is applied o Decision-Making suppo is analysed in Sec ion 6. In Sec ion 7, he de ails o he DSS ha we p o- pose, he implemen ed applica ion o include BDCs in he decision-making suppo , and a case o s udy a e explained. Sec ion 8 discusses p e ious wo k ela ed o ou p oposal. Finally, conclusions a e d awn and u u e wo k is p esen ed. 2. Rep esen a ion o he Decision-Making Objec i es As men ioned abo e, BDCs a e a subse o he business compliance ules, o ien ed o da a low alues. BDCs can be used o ep esen he ela ion be ween he alues o he a iables ha low in a business p ocess ins ance, by helping in he desc ip ion o he policy o he company and in he decision-making suppo o inpu da a. Al hough he e a e se e al language cons uc s o he design o business compli- ance ules, mos a e based on he use o IF-THEN ules o hei de i ed ex ensions o ECA ules (e en -condi ion-ac ion ules) o ECAA ules (e en -condi ion-ac ion- al e na i e ules). The languages o ep esen ing business ules a y be ween e- sea ch p o o ypes (e.g. N3), endo speci ic o ma s (e.g. D ools, Fai Isaac Blaze Ad iso , ILOG JRules and Jess), and p oposals o he XML-based exchange o business ules (e.g. SRML, PRR, and SBVR). Ano he possibili y is he Business P ocess Compliance Language (BPCL) 8, which de ines inclusion, p ecedence, and exis ence condi ions o business ules by means o Objec Cons ain Language (OCL) exp essions, which speci y co ec ness and compliance checks. Tha e sion o BPCL was imp o ed in 6 o de elop a seman ic app oach o business ule man- agemen ha allows in ui i e modelling and analysis o business p ocess compliance. In 7, a g amma o BDCs is p esen ed. In his pape , we ex end his g amma o include new ope a o s ha can acili a e he desc ip ion o he policy o he com- panies. This ex ension is inspi ed by he idea o iewing a BDC as a Cons ain : a Boolean combina ion o equa ions and inequa ions ha ollows he me amodel o Figu e 2 based on ha p esen ed in 9. A BDC can be an a omic cons ain , a nega ion o a cons ain , o a bina y cons ain o med by wo cons ain s joined wi h a Boolean ope a o (AND, OR, IMPLY). An a omic cons ain is o med by wo unc ions and a compa a o (<, ≤, ≥, . . .). Each unc ion can be a una y unc ion (a a iable), a cons an , o a bina y unc ion joined wi h an ope a o (+, ∗, −, /). The me amodel is based on he ecu si e de ini ion o Cons ain , since wo cons ain s can be combined in o a new one, by using one o he Logic Ope a o s o a Nega ion. Fo example, he cons ain s: {a+b ≤ c ∧ c > 8} and {a*b ≥ 5 ∨ ¬(c ≤ 15)}, can be combined in o a new cons ain , o example by using he ∨ ope a o , he eby ob aining he new cons ain : {(a+b ≤ c ∧ c > 8) ∨ (a*b ≥ 5 ∨ ¬(c ≤ 15))}. We he e o e de ine ou me hods o he combina ion o a BDC (c) wi h ano he (c’): •c.and(BDC c’) e u ns a new cons ain : {c∧c’} •c.o (BDC c’) e u ns a new cons ain : {c∨c’} •c.no () e u ns a new cons ain : {¬ c} •c.imply(BDC c’) a new cons ain : {c→c’} These me hods will be used in Algo i hm 5.2 o build he BDC ha will ep esen all possible alues o a a iable in acco dance wi h he model o he business p ocess. Cons ain - domain : Va iableType - name : S ing Va iable - FLOAT - INTEGER - NATURAL <<enum>> Va iableType - isDe inedBy *1 - NOT_EQUAL - EQUAL - GREATER_EQUAL - GREATER - LESS_EQUAL - LESS <<enum>> Compa a o - compa a o : Compa a o A omicCons ain Nega ionCons ain - booleanOp : LogicOpe a o Bina yCons ain - IMPLY - OR - AND <<enum>> LogicOpe a o Func ion - isFo medBy 2 1 Una yFunc ion - has 1 - alue : S ing Cons an - uncOpe a o : Func ionOpe a o Bina yFunc ion - DIVIDE - MULTIPLY - MINUS - PLUS <<enum>> Func ionOpe a o - has 2 1 - ela es 2 - nega es 1 Fig. 2. Cons ain Me amodel 3. A mo i a ing example The mo i a ing example used in his pape is he well-known example o he o ga- niza ion o a con e ence, o which a educed model is p esen ed in Figu e 3. This business p ocess shows an example whe e decisions abou inpu da a mus be made a a ious poin s o he business p ocess, whe e he u u e alues o se e al a i- ables emain unknown, since hey a e in oduced in o ac i i ies ha ha e ye o be execu ed. Fi s o all, he o ganizing commi ee has o de e mine he ea ly and la e egis a ion ees se e al mon hs be o e he numbe o pa icipan s is known, and his decision canno be changed a e he call o pape s has been made. A simila si ua ion exis s when deciding he numbe o p oceedings ha will be p in ed, bu al hough he inal numbe o pa icipan s emains unknown un il he con e ence ends, he e is a signi ican ela ionship be ween he numbe o accep ed pape s and he numbe o pa icipan s. O he decisions, such as which es au an o book o he gala dinne and lunches, ha e o be made al hough he numbe o pa icipan s can in luence he de e mina ion o he p ice and he choice o es au an . Al hough his in o ma ion is unknown, i is necessa y o make many decisions be o e he con e ence s a s. In he example, nine di e en inpu da a a iables pa icipa e in he da a decision-making p ocess (ea ly egis a ion ee, la e egis a ion ee, lunch p ice, dinne p ice, numbe o p oceedings o p in , local speake cos , in e na ional speake cos , publici y and enue cos ). The e a e o he inpu a iables, bu hei Es ablish Con e ence Ra e Con ac Pa ne P in P oceedings Final epo and make paymen s In i e Local Speake In i e In e na ional Speake Au ho No i ica ion Regsi e Ea ly Income- o alCos <4000 / Book Dinne Book Lunch Pape Submission Sen Pape Accep ed Pape Num Ea ly egis a ions Num La e Regis a ions Hold Con e ence Sponso ship AND 1 AND 2 XOR1 OR 1 Ea ly Regis a ion Fees La e Regis a ion Fees Venue Cos publici y P oceeding P ice Numbe O P oceedings Con e ence lunch p ice Con e ence dinne p ice sponso ship>=5000 sponso ship>=8000 in e Speake Cos localSpeake Cos Fig. 3. Example o a con e ence o ganiza ion p ocess alues a e de e mined in a manda o y way, o example he numbe o accep ed pa- pe s, o he inal numbe o pa icipan s ha belong as pa o he da a low. Since decisions ha e o be made by he o ganizing commi ee, wo ques ions a ise: Would i be mo e conduci e owa ds he success o he con e ence o ake in o accoun all he po en ial ac s in he u u e, and he possible b anches ha will be execu ed depending on he speci ic alues o each p ocess ins an ia ion? How i can be done? Ob iously, i he e is no in o ma ion abou he ela ionship be ween he a i- ables ea lyRegis a ionFee,la eRegis a ionFee,lunchP ice, e c, no ype o in e ence abou he possible co ec alues can be aken. Fo his eason, BDCs a e necessa y. Examples include: • {sen Pape *0.3 ≤accep edPape s ≤sen Pape *0.8}associa ed wi h he ac i i y Pape Submission. • {numEa lyPa icipan *0.8 ≤accep edPape s ≤numEa lyPa icipan * 1.2}associa ed wi h he ac i i y Au ho No i ica ion. • {numO P oceedings ≥1.1 * numbe O Pa icipan s}associa ed wi h he ac i i y P in P oceedings. • {Income*0.80 ≤To alCos ≤Income*0.9}associa ed wi h he ac i i y Hold Con e ence. These BDCs, combined wi h he p ocess model and he condi ional sequence lows p esen ed in Figu e 3, can be applied in o de o asce ain he decision a i- ables, o example lunchP ice in he ac i i y Book Lunch. All he a iables and BDCs a e explained in de ail in he Appendix o he pape . 4. Fo maliza ion o he Decision-making suppo o inpu da a in Business P ocesses In his sec ion, basic no a ion and concep s a e in oduced and he o malism used o exp ess ou p oblem is b ie ly desc ibed. A DSS o inpu da a in business p o- cesses unde akes asks such as an e alua ion o al e na i es, and communica es i s conclusions by aking in o accoun he BDCs ha mus be consis en in u u e decisions and in he p ocess model. The e o e, he same model can be in ol ed a di e en poin s o decision-making acco ding o he decision a iables ha he use needs o asce ain, and o he speci ic alues ins an ia ed o he a iables o he da a low. The e o e, wo aspec s a e combined in he decision-making suppo : he model M(<P,BDC,DF>), o med by ac i i ies and con ol lows, he BDCs associ- a ed o each ac i i y, and he da a low a iables; and he decision poin DP, o med by he decision a iables (whose possible alues need o be asce ained) and all he ins ances o he da a low a iables un il he momen <DV,DFI>. Acco ding o hese wo desc ip o s, he decision-making suppo can be pe o med by ob aining all he possible DVI uples o alues o DV ha sa is y DBC and he DFI. A DSS o he inpu da a o a business p ocess can he e o e be speci ied by means o he Decision P ocess Model: De ini ion o Decision P ocess Model: This is o med by means o he business p ocess model (M), and he decision poin s (DP). Each o hese pa s a e de ined a he same ime as: M=<P,BDC,DF> DP =<DV,DFI> By using hese wo pa s o he desc ip ion, he decision-making suppo o inpu da a ob ains he DVI, which is a se o co ec uples o ins an ia ion o DV: <M,DP>→DVI | ∀ di∈DVI,{BDC ∪DFI ∪ di}`> The ollowing subsec ions o malize he model, he decision poin s, and speci y how o ob ain he decision-making suppo in de ail. 4.1. Business P ocess Model (M) A pa o he Decision P ocess Model is he p ocess model P, and is composed o : •SE, one s a e en o ini ialize he p ocess. •EE, a se o end e en s, wi h a leas one elemen . •A, a se o ac i i ies ha de ines he model o he p ocess. •CF, a se o con ol low pa e ns (AND, OR, XOR) ha desc ibes he possible b anches o execu e. •Cond, a se o condi ions associa ed wi h he con ol lows OR and XOR, ha desc ibes he pa hs ha he p ocess can ake depending on he alues o he a iables in he da a low. These condi ions a e e alua ed a un ime, A1 A2 A3 A4 BDC1 BDC2 BDC3 BDC4 BDC5 BDC6 BDC7 Ac i i ies Business Da a Cons ain s Fig. 4. Func ion ela ion be ween Ac i i ies and BDCs when he alues o he a iables a e known. The BDCs associa ed o each ac i i y o he p ocess (BDC) a e de ined wi h he da a low a iables DF; he DF can hen be de ined o a ange o possible alues o assis in he decision suppo (DFR). The ela ionship be ween ac i i ies and BDCs is p esen ed in Figu e 4, whe e he e is a su jec i e co espondence be ween he Ac i i y and he Business Da a Cons ain se s. The impo ance o he in eg a ion o business ules and business p ocesses was analysed in 33. E e y BDCiin a Business Da a Cons ain se has a co esponding elemen Ajin he Ac i i y se , such ha (Aj) = BDCi, whe eby mul iple ac i i ies migh be u ned in o he same BDC by applying , al hough no all he Ac i i y elemen s mus ha e a ela ion wi h he elemen s o he Business Da a Cons ain se . The e o e, he BDC is desc ibed by means o he uple: BDC =<DFR, :A→BDCs>. The BDCs ha desc ibe he ange o he possible alues o each da a low a i- able a e assigned o he whole p ocess. Fo he p ocess example p esen ed in Figu e 3, he componen s o he p ocess acco ding he de ini ions abo e a e: •A={Es ablish Con e ence Ra e, Con ac Con ibu ion, Pape submission, Au ho No i ica ion, ...}. •CF ={AND1, AND2, XOR1, OR1}. •Cond o he con ol low pa e n XOR1is {Income- o alCos <4000}, and o he con ol low pa e n OR1a e {sponso ship ≥5000, sponso ship ≥ 8000}. •BDC is composed o : –DFR ={5000 ≤ o alCos ≤50000},{5000 ≤income ≤60000},{50 ≤numEa lyPa icipan s ≤200}, ... – :A→BDCs =Es ablish Con e ence Ra e→ {1.2 * ea lyReg- is a ionFee ≤la eRegis a ionFees ≤1.5 * ea lyRegis a ionFee, ix- Algo i hm 5.2 Recu si e Algo i hm o a p ocess g aph 1: unc ion G aphT a e sal(G aph g, Node n, Cons ain c) 2: while n is no an END node do 3: i n is an Ac i i y hen 4: i c == new Cons ain () hen 5: c = n.ob ainCons ain (); 6: else 7: c.and(n.ob ainCons ain ()); 8: end i 9: n = g.ob ainNeighbou (n); 10: .A single node is e u ned ( o line 3 o P ocess G aph de ini ion). 11: else i n. ype is a Spli Con ol Flow hen 12: Se nodes = g.ob ainNeighbou (n); 13: .Se e al neighbou s a e ob ained, one o each b anch. 14: Node n1 = nodes.ge (); 15: Cons ain c1 = new Cons ain (); 16: A ay A ayNodes[] = new Nodes[nodes.size()]; 17: A ayNodes[0] = G aphT a e sal(g, n1, c1); 18: c1.add(g.label(n,n1)); 19: in ege i = 1; 20: while nodes.nex () do 21: Node n2 = nodes.ge (); 22: Cons ain c2 = new Cons ain (); 23: A ayNodes[i++] = G aphT a e sal(g, n2, c2); 24: c2.and(g.label(n, n2)); 25: i n is an OR o an XOR con ol low hen 26: c1.OR(c2); 27: else 28: c1.AND(c2) 29: end i 30: end while 31: c.and(c1); 32: n = heNodeDis inc O End(A ayNodes); 33: else .Any join con ol low 34: n = g.ob ainNeighbou s(n); 35: .A single node is e u ned ( o line 2 o P ocess G aph de ini ion). 36: e u n n; 37: end i 38: end while 39: e u n n 40: end unc ion Final epo and make paymen s Any BDCs de ined o he whole p ocess, and no jus o a speci ic ac i i y, will be included wi h an AND Boolean ela ion wi h he BDCs ob ained om he execu ion o he algo i hm. I he e is an ac i i y wi hou any associa ed BDC, i will be equal o no including a cons ain o including a ue cons ain in he BDC ob ained. 6. E alua ing he Decision P ocess Model a he Decision Poin s As men ioned in Sec ion 4, an impo an key o he decision-making suppo o inpu da a is how o p esen he in o ma ion so ha i is use ul o he use . In o de o ob ain a nume ical ep esen a ion by means o in e als, we p opose he use o he Cons ain P og amming pa adigm o assu e ha he decision model can be e alua ed in an e icien way. This assu ance is hanks o he CSP o mal ep esen a ion being e y simila o he o mal ep esen a ion o he decision model p esen ed in his pape . The e o e, we p opose modelling and e alua ing a Con- s ain Sa is ac ion P oblem wi h he BDC ob ained om Algo i hm 5.2, and o he ins an ia ed a iables o da a low. A Cons ain Sa is ac ion P oblem (CSP) ep esen s a easoning amewo k con- sis ing o a iables, domains and cons ain s. Fo mally, i is de ined as a uple <X, D,C>, whe e X={x1,x2,. . .,xn}is a ini e se o a iables, D={d(x1), d(x2), . . .,d(xn)}is a se o domains o he alues o he a iables, and C={C1,C2, . . .,Cm}is a se o cons ain s. Each cons ain Ciis de ined as a ela ion Ron a subse o a iables V={xi,xj,. . .,xl}, called he cons ain scope. The ela ion Rmay be ep esen ed as a subse o he Ca esian p oduc d(xi)×d(xj)×. . . × d(xl). A cons ain Ci= (Vi,Ri) simul aneously speci ies he possible alues o he a iables in Vin o de o sa is y R. Le Vk={xk1,xk2,. . .,xkl}be a subse o X, and an l- uple (xk1,xk2,. . .,xkl) om d(xk1), d(xk2), . . .,d(xkl) can he e o e be called an ins an ia ion o he a iables in Vk. An ins an ia ion is a solu ion i and only i i sa is ies he cons ain s C. In o de o sol e a CSP, a combina ion o sea ch and consis ency echniques is commonly used 11. The consis ency echniques emo e inconsis en alues om he domains o he a iables du ing o be o e he sea ch. Du ing he sea ch, a p opaga- ion p ocess is execu ed which analyses he combina ion o alues o a iables whe e he cons ain s a e sa is iable. Se e al local consis ency and op imiza ion echniques ha e been p oposed as ways o imp o ing he e iciency o sea ch algo i hms. In a CSP, he inclusion o a cons ain in he se Chas he same e ec as including his cons ain wi h an ∧ ela ion wi h he se C. Fo his eason, in his case he CSP will be composed o he a iables o he da a low, bo h ins an ia ed and non-ins an ia ed, o he BDC ob ained om Algo i hm 5.2, and o he BDCs de ined o he whole p ocess. The pa s o he CSP acco ding o he de ini ion o Decision P ocess Model a e he e o e: •X:DF •D:DFI •C:{BDCs de ined o he whole p ocess} ∪ {BDC ob ained om he execu ion o he algo i hm G aphT a e sal} Since he CSP e u ns all he possible alues o he a iables (DF in his case), i is necessa y o educe i o p esen only he alues o he decision a iables (DV). To his end, he decision a iables a e de ined as objec i es du ing he p opaga ion p ocess whe e he a iables a e ins an ia ed. This enables he sea ch o s op he ins an ia ion in he b anches whe e no new alues o decision a iables can be ound, he eby bounding he unnecessa y combina ions o alues. Fo each solu ion ound, each alue o he decision a iables is s o ed in a so ed lis . Each o hese so ed lis s is ea ed in o de o e u n he lis o in e als o each a iable o decision. Fo example i he alues {1, 2, 3, 5, 8, 9, 10} a e ound o he a iable x, he lis o in e als buil is {[1, 3], [5, 5], [8, 10]}. Fo he example o Figu e 3, he CSP buil o analyse he possible alid alues o he a iable o decision Numbe O P oceedings in he ac i i y P in P oceeding (which uses he a iables and BDCs p esen ed in he Appendix) is: //All he a iables o he da a low o alCos , numEa lyPa icipan , numLa ePa icipan , ... In ege //The a iables ins an ia ed un il he decision poin Ea lyRegis a ionFee = 500 La eRegis a ionFee = 750 .. . //Range o he da a low a iables and BDCs o he whole p ocess o alCos [5000..50000]In ege numEa lyPa icipan s[50..200]In ege numLa ePa icipan s[10..100]In ege numPa icipan s[60..300]In ege Cos Pe Pa icipan = 3*lunchP ice+dinne P ice+p oceedingsP ice .. . //BDCs ob ained om he algo i hm G aphT a e sal BDCs o he Ac i i y Es ablish Con e ence Ra e BDCs o he Ac i i y Con ac Pa ne ... (((Income- o alCos <4000) ∧( ue)) ∨(¬(Income- o alCos <4000) ∧ (sponso ship≥5000)∧(Local in i a ion) ∨ ((sponso ship≥8000) ∧(In e na ional in i a ion)))) //BDCs o he Ac i i y Hold Con e ence //BDCs o he Ac i i y Final epo and make paymen s Goal o b anching(numO P oceedings) The CSP sol e used in ou p oposal is ChocoT M 12. Once he esolu ion o he CSP has inished, he lis o in e als ob ained is used o in o m he use abou he possible co ec alues. Fo he example, once ea lyRegis a ionFees,la eRegis a- ionFees, enueCos ,sponso ship,sen Pape s and accep edPape s ha e been ins an- ia ed in he p ocess wi h he alues {500, 750, 2500, 8500, 150} espec i ely, in he ac i i y P in P oceedings he in e al {[67, 123]}is ob ained in o de o asce ain P ocess Laye Business Da a Cons ain s Reposi o y P esen a ion Laye Applica ion Laye Pe sis ence Laye Da a Inpu Decision Making Suppo Laye Nume ical Sol e Symbolic Sol e Fig. 7. DSS o Decision-making suppo o Inpu Da a he possible alues o he decision a iable numbe O P oceedings. 7. Implemen a ion de ails in a Case o S udy In o de o show he bene i s o he decision-making suppo o inpu da a in business p ocess ins ances, we ha e implemen ed a solu ion based on he P ocess Awa e In o ma ion Sys em (PAIS) amewo k and illus a ed how o acili a e he inpu da a suppo in o a comme cial solu ion. 7.1. Decision Suppo Sys em o Inpu Da a in Business P ocesses In o de o pe mi he decision-making suppo o inpu da a a a ious poin s o he p ocess based on he BDCs, his pape is based on an ex ension o he classic PAIS amewo k 16, as p esen ed in 17, and shown in Figu e 7. In gene al, a PAIS a chi ec u e 18 can be iewed as a 4- ie sys em as p esen ed in 16, whe e, om op o bo om, he laye s a e: P esen a ion Laye , P ocess Laye , Applica ion Laye and Pe sis ency Laye . As a undamen al cha ac e is ic, PAIS p o ides he means o sepa a e p ocess logic om applica ion code. Da a decision-making suppo and business p ocess laye s a e wo pa allel and ”independen ” sys ems. They a e conside ed independen since hey can be simul- aneously execu ed in sepa a e machines, o di e en applica ions. Howe e , his independence ails om he poin o iew o da a low in o ma ion, since, o he a ious decision poin s, he Da a Inpu Decision laye uses he ins an ia ed a i- ables in he da a low and uses he decision a iables ha he p ocess needs. Wi h he p esen ed DSS, i is possible o design bo h he business p ocess model and he BDCs, he eby achie ing highe le els o lexibili y and agili y in he business p ocess managemen . One o he i ems ha also needs o be s udied is how o s o e he BDCs, and how o de ine he ela ion be ween each o hem and he ac i i ies o he p ocess. When a g ea deal o BDCs ha e o be handled, he use o a da abase o s o e and manage hese cons ain s is manda o y, especially when no all he BDCs a e es ablished o he whole business p ocess, and he ela ion be ween ac i i ies and BDCs has o be de ined. The necessi y o s o e he business compliance ules was analysed in 19, bu ailed o ake da a seman ics in o accoun . Howe e , BDCs canno be s o ed in a classic ela ional da abase, since s o ing a BDC also implies s o ing all he de- ails ela ed o i s a iables, he domain o a iables, and da a pe sis ence ela ion- ships. The di icul y in s o ing BDCs a ises due o he p oblem o how o s o e he cons ain s hemsel es as da a, since hey do no belong o a ype suppo ed by comme cial da abases. In o de o manage cons ain s, we p opose he use o Con- s ain Da abase Managemen Sys ems (CDBMS) as explained in 20. Tha p oposal is based on an en elope o a da abase managemen sys em o manage Cons ain s as a classic ype, which has been p oposed o he desc ip ion o he BDCs. A sim- ila way o s o e BDCs was used in p e ious wo k 7. This solu ion shields he use om unnecessa y de ails on how he BDCs a e s o ed and que ied. Once how o s o e he BDCs is asce ained, he nex ques ion is how each BDC is ela ed o each ac i i y and wi h he es o he p ocess. Figu e 8 ep esen s he ela ions necessa y o desc ibe ha a P ocess has a se o Da a low a iables, a ailable o he Ac i i ies. Each BusinessDa aCons ain can be associa ed wi h a se o ac i i ies o o he whole p ocess, while each Ac i i y can ha e se e al associa ed BDCs. These associa ions a e es ablished in he able Ac i i y/BDC. Ac i i ies (pk) IdAc i i y: in Name: S ing Ac i i y/BDC (pk) IdBR: in (pk) IdAc i i y: in 1..1 0..n Business Da a Cons ain s (pk) IdBDC: in ule: Cons ain 0..n 1..1 Business P ocesses (pk) IdP ocess: in Name: S ing bpmn2.0: XML Da aFlowVa s (pk) IdVa iable: in Name: S ing Type: S ing 1..1 0..n 1..1 1..n 0..n 0..1 Fig. 8. Rela ions be ween BDCs and Ac i i ies P esen a ion Laye } Connec o MARTIN (MAking Reasoning o daTa INpu ) P ocess Laye Da a Inpu Decision-Making Suppo Laye 1 4 2 3 Fig. 9. DSS o he decision-making suppo o inpu da a wi h ools 7.2. Compu a ional applica ion o he Decision-Making suppo o Inpu Da a In o de o acili a e he c ea ion o BDCs and he decision-making o he p ocess, we ha e implemen ed an applica ion and a connec o ha ollow he DSS p esen ed in Sec ion 7.1. This solu ion uses a speci ic se o echnologies ha could be eplaced by ano he se . The speci ic con igu a ion ha we ha e implemen ed is p esen ed in Figu e 9, which desc ibes a possible combina ion o ools o execu e he decision- making suppo o inpu da a. We ha e p epa ed a ideo 21 whe e he s eps o he design and execu ion o he decision suppo a e shown. The s eps o con igu e and use he applica ion a e: (1) Modelling he business p ocess and de ining he da a low a iables (P, DF): The business p ocess and he da a low a iables can be modelled in any Business P ocess Managemen Sys em, o example: In alioT M , Ac i i iT M , and Boni a Open Solu ionT M . In he case o s udy p esen ed, we ha e used Boni a Open Solu ionT M since i is an open-code applica ion wi h ee dis ibu ion, and is commonly used in he p i a e company sec o . Once he p ocess is modelled, he designe o he p ocess mus decide on he decision poin s associa ed wi h any ac i i y. (2) Loca ing he decision poin s and decision a iables (DV): I a designe conside s alloca ing a decision poin in o a de e mined ac i i y, hen a connec- o mus be added in he ac i i y o ela e i wi h he so wa e ha execu es he decision. We ha e implemen ed a connec o (as shown in he ideo 21) o acili a e he ela ion be ween he decision poin and he easoning so wa e (MARTIN) ha ob ains he possible and co ec alues o he decision a i- ables. When a connec o is included in an ac i i y, i is necessa y o de ine he decision a iables. In he connec o , called ”Da a Decision-Making Connec o ”, all he da a low alues and he decision da a a iables a e sen a un ime o he so wa e ha execu es he e alua ion o he decision-making suppo o he model ob ained a design ime, as explained in Sec ion 6. (3) C ea ing he BDCs (BDC): Once he p ocess model is de ined, i is necessa y o c ea e he BDCs and associa e hem o each ac i i y, using he solu ion o Figu e 8. The decision-making p ocess uses all he a iables ins an ia ed in he da a low a un ime, and he BDCs ha ep esen he possible co ec alues in he u u e. To his end, we ha e implemen ed an applica ion called MARTIN: MAking Reasoning o daTa INpu , o acili a e he c ea ion o BDCs, and he associa ion o he BDCs o he ac i i ies. Mos comme cial ools p o ide an XML ep esen a ion o he c ea ed p ocess ha ollows he BPMN 2.0 13. Fo his eason, we ha e implemen ed a ans o ma ion om a .bpmn ile ha ep e- sen s he ype models desc ibed in his pape ( o med by ac i i ies and con ol lows), in o ou P ocess G aph. When he XML o he p ocess is analysed, he ables Business P ocess,Ac i i ies, and Da a lowVa shown in Figu e 8 a e illed in au oma ically, since he XML o he p ocess holds all his in o ma ion. I is hen possible o c ea e and/o assign BDCs o he a ious ac i i ies. In o de o p o ide a simple way o he business expe o add BDCs, he in e ace (shown in Figu e 10.a) enables he business p ocess model o be iewed, and BDCs o be assigned o he ac i i ies ha belong o he business p ocess (Figu e 10.b). a) C ea ing BDCs b) Assigning BDCs Fig. 10. Connec o o c ea e and assign BDCs o ac i i ies (4) Ins an ia ing he business p ocess model (DFI) and ob aining he de- cision a iables ins an ia ion (DVI). Once he business p ocess model has been designed in a Business P ocess Managemen Sys em, he execu ion p o- cess can be pe o med. When a business p ocess ins ance execu es an ac i i y wi h a Da a Decision-Making Connec o , he da a low alues o ha ins ance and da a decision a e sen o he Da a Inpu Decision-Making Suppo laye which implemen s he e alua ion o models, as explained in Sec ion 6. Figu e 11 p esen s an example o he o m ha shows he possible in e als o each decision a iable. Fig. 11. Fo m o Ac i i y Es ablish Con e ence Ra e a un ime 8. Rela ed Wo k Decision-making suppo in business p ocesses ca ies signi ican con ibu ions e- la ed o how o model he p ocess, which in u n help he designe o decide he bes combina ion o ac i i ies o achie e an objec i e. A simula ion-based app oach o decision-making suppo is p oposed in 22 wi h espec o complex dynamic sys- ems, and includes unce ain da a. A me hodology o op imize a p ocess whe e he desc ip ion is no clea ( uzzy) is pu o wa d in 23. Howe e , in bo h pape s, he help in he business p ocess has been o ien ed owa ds he design o he model, o he edesign o he business p ocess 24, by looking a he quali y o he p ocess a design ime, bu no how his p ocess wo ks a un ime, he eby missing he impo ance o he a iables o he da a low. Da a has also been in ol ed in o he s udies ela ed o decision suppo ; o example 25 p oposes an ope a ional decision suppo o he cons uc ion o p ocess models based on his o ical da a o simula e p ocesses. Tha p oposal includes a gene ic app oach o a business p ocess o op- e a ional decision suppo , and includes business p ocess modelling and wo k low simula ion wi h he models gene a ed, by using p ocess mining. O he wo k ela ed o how o model he p ocesses, such as ha in 26, p oposes a amewo k o assis- ance o c ea e models which ake he necessa y esou ces in ol ed in he p ocess in o accoun . In ha pape , he da a ha desc ibe he esou ces o he execu ion o he p ocess a e used, bu no he da a ha lows a un ime, no does i conside how his assis ance can help a un ime. In gene al, pape s ound in he li e a u e ela ed o decision-making suppo a e no ocused on he assis ance o he use o he inpu da a. Al hough wo k such as 27 is o ien ed owa ds audi ing he p ocess in o de o de ec gaps be ween he in o ma ion sys em p ocess low and he in e nal con ol low in he business p ocess, he quali y o he da a alues a un ime is no a cause o conce n o he au ho s. This de ec ion can be de i ed om he exis ence o an o e sigh in he desc ip ion o he seman ics o da a in he business p ocesses. The use o compliance ules has adi ionally been used o he alida ion o he business p ocess, no o use assis ance. The alida ion o business p ocess aces has been a ield o in ense esea ch o e ecen yea s using business compliance ules: see 28 as an en y poin in o his li e a u e. Howe e , hese ypes o p oposal canno be used in he decision- making suppo o inpu da a, since hey a e ocused on he compliance o he p ocess model s uc u e 29, 30. Rega ding how o model da a-awa e compliance ules, s udies such as 10, 31, 32, and 34, ha e de ined g aphical no a ions o ep esen he ela ionship be ween da a and compliance ules by means o da a condi ions. These ypes o compliance ules canno he e o e be used o in e he possible alues o he a iables ha a e in ol ed in he decisions. In 35, ”seman ic cons ain s” and he SeaFlows amewo k o enabling in eg a ed compliance suppo a e p oposed. Fu he mo e, in 37, a p ep ocessing s ep o enable da a-awa e compliance checking in an e icien manne is p esen ed: he da a desc ibe unde wha condi ions he ac i i ies can be execu ed. In gene al, many examples can be ound whe e da a objec s a e used o compliance e i ica ion, o ins ance, he seman ically anno a ing ac i i ies wi h p econdi ions and e ec s ha may e e o da a objec s a e in oduced in 38, bu none o hese ac o s assis he use wi h his in o ma ion a un ime. Summa izing, o he bes o ou knowledge, only one p elimina y s udy 17 exis s ha uses he knowledge o he business p ocess model and he BDCs o decision- making suppo o inpu da a, while he es o he p oposals a e ocused on he design o e-design o he business p ocess model. The cu en pape cons i u es an imp o emen on he p e ious pape by: including he p ocess g aph de ini ion o ep esen he model; de ining he algo i hm o a e se he g aph; and by imple- men ing an applica ion ha can be in eg a ed wi h comme cial ools. 9. Conclusions and Fu u e wo k In his pape , he use o BDCs is p oposed o in e he possible inpu da a alues in a business p ocess ins ance, by aking in o accoun he model o he p ocess and he decision poin s whe e he assis ance mus be execu ed. In o de o mee his chal- lenge, wo di e en pa s ha e been dis inguished: an analysis o he p ocess model a design- ime, by means o an algo i hm o a e se he business p ocess model; and an e alua ion o he decision model a un ime, o ob ain he possible alues o he decision a iables using cons ain p og amming. In o de o implemen his p oposal, a case o s udy has been de eloped using Boni a So and an applica ion called MARTIN has been included in o de o c ea e, include, and e alua e he BDCs in he business p ocess model. One aspec ha can be imp o ed in his wo k is he p esen a ion o he possible and co ec alues o he decision a iables o he use . In he case o nume ical decision suppo , his imp o emen could be a ained by showing he use he pos- sible combina ions o a ious decision a iables simul aneously, ins ead o solely p esen ing he in e als o each a iable. As u u e wo k, we also plan o include a symbolic ep esen a ion o he possible co ec a iables ins ead o in e al alues. Fu he esea ch could be analysed in g ea e dep h, such as: a) In he cu en p oposal, all he BDCs a e de ined o all ins ances, bu i would be possible, de- pending on he alues o he da a low o he momen whe e he ins ance is execu ed, ha a di e en se o BCDs could be in ol ed in he decision suppo . b) Rela ed o he abo e p oposal, he e is also he possibili y o including new BDCs o he use a un ime, which would ende he decision-making suppo o inpu da a mo e cus omized. This would help he use o include, o example, educ ions o he domains o de e mined decision a iables associa ed wi h u u e decision poin s, e en be o e he inal alue is in oduced. I , a a decision poin , he e a e no possi- ble alues o a decision a iable, i implies ha no BDC is consis en . Fo u u e wo k, we p opose ein o cing he de ec ion o he minimum se o BDCs ha a e no sa is iable, and asce aining how his si ua ion can be ixed by means o inpu da a decisions. Acknowledgemen This wo k has been pa ially unded by he Jun a de Andaluc´ıa by means o la Conseje ´ıa de Inno aci´on, Ciencia y Emp esa (P08-TIC-04095) and by he Minis y o Science and Technology o Spain (TIN2009-13714) and he Eu opean Regional De elopmen Fund (ERDF/FEDER). Appendix Da a low Va iables: • o alCos ,income is he o al cos and income o he con e ence espec i ely •numEa lyPa icipan s is he numbe o people a ending, egis e ed in he ea ly pe iod •numLa ePa icipan s is he numbe o people a ending, egis e ed in he la e pe iod •numPa icipan s is he inal numbe o pa icipan s •cos Pe Pa icipan is he cos pe pa icipan , and includes lunch and dinne , cos o p oceedings, e c. •ea lyRegis a ionFee is he cos o ea ly egis a ion •la eRegis a ionFee is he cos o la e egis a ion •sponso ship is he income om he companies ha sponso he e en •dinne P ice is he p ice o he gala dinne