Organization-Driven Business Process Configurable Modeling for Spatial Variability

Guosheng Kang,Liqing Yang,Liang Zhang,*,Jianxun Liu,Yiping Wen

1Hunan Provincial Key Lab.for Services Computing and Novel Software Technology,Hunan University of Science and Technology,Xiangtan 411201,China

2School of Computer Science and Engineering,Hunan University of Science and Technology,Xiangtan 411201,China

3School of Computer Science,Fudan University,Shanghai 201203,China

Abstract:Nowadays,enterprises need to continually adjust their business processes to adapt to the changes of business environments,especially when one business needs to be deployed in different application scenarios,which is called spatial variability in this paper.In the field of BPM(Business Process Management),configurable business process models have demonstrated their effectiveness in aspects of process modeling and model reuse.Yet,we found that the existing techniques lead to complex configurable models,and are inadequate for model reuse especially for the spatial variability issue because they neglect the root impact of organizations on control flow.S-BPM(Subject-oriented Business Process Management)models provide a solid foundation for dealing with complex applications and help to bridge the gap between business and IT for process execution.In this paper,we propose an organization-driven business process configurable modeling approach for spatial variability by integrating both restriction and extension operations based on the S-BPM paradigm,in which business objects are also included.Our approach is validated with a general business process developed for the Real Estate Administration(REA)in a certain province of China.The resulting configurable modeling framework can express the heterogeneous activity sequences for one business and has the potential to generate process models for uncertain environments in a new organization structure.

Keywords:S-BPM;configurable modeling;organization-driven;spatial variability;business objects;services

I.INTRODUCTION

Today’s enterprises are situated in a dynamic world where new risks but also opportunities,often suddenly occur.The organizations which can react quickly to the changing environment have a strong competitive advantage.Variability issues in business processes(BPs)are common,especially when one business needs to be deployed in different application scenarios,which is calledspatial variabilityin this paper.For a typical example,to enlarge the scale of a business,one enterprise has to deploy its BPs in different cities.However,different cities have different organization structures,government regulations,and other different factors.Such differences ask for different process models for a business.To cope with the changes in today’s business requirements and reduce the cost of process modeling,an effective method to support spatial variability modeling in different application scenarios turns into a pressing need,which is the motivation of this paper.

Currently,many approaches are proposed to deal with the variability of BPs.They are motivated by the“Design by Reuse”paradigm.They can be classified into two categories:one is variability by restriction,and the other is variability by extension[1,2].Variability by restriction starts with a configurable process model that contains the behaviour of all process variants.Configuration is achieved by restricting the behaviour of a configurable process model.Then individualization is performed to derive a specific process variant.A configurable process model guides the users to a solution that better fits a specific working context.The typical representatives are languages such as C-EPC[3],C-iEPC[4],vBPMN[5],or CYAWL[6].Also,language-independent approaches have been proposed.However,variability by extension takes the opposite starting point.The customizable process model represents the most common behaviour that is shared by most variants.At customization time,the model’s behaviour needs to be extended to serve a particular situation.The notable examples are PROVOP[7]and BPFM[8].

However,we found that the existing techniques are inadequate for model reuse due to the following reasons:(1)The existing techniques dealing with variability focus on temporal variability while spatial variability is neglected.For example,when one business needs to be deployed in a new city,the model is hard to reuse due to the very different organization structures,legislations,and authorizations.Thus,spatial variability becomes a research issue.(2)Most existing approaches focus on control flow directly,neglecting the aspect of organization structures.From our investigation,in many real cases in China,organization structures have a direct impact on activities.Thus,which tasks of a business are required is decided by whether the corresponding departments exist and have sufficient delegation and capability,which is calledorganization dependent contextin this paper.Therefore,the required organization departments should be considered before modeling tasks[3–8].(3)The existing techniques focus on modeling process orchestration,leading to complex configurable models(sometimes with redundant tasks).Designers are in dilemma to do the configuration for desirable results.Thus,the complexity of modeling should be relieved for modelers.(4)The existing approaches neglect the details of data information.Either data is not considered or it is considered as a data object neglecting the inner structure of the data model[4].The under specification of data hardens the implementation of a process model since different data information is required in different cities.In this paper,we aim to solve the spatial variability issue of business process modeling under organization dependent context.

Subject-oriented Business Process Management(S-BPM)is a methodology for modeling business processes.S-BPM models can be directly interpreted by an appropriate work flow engine[9]and the simple semantics of the modeling language(consisting of five symbols)meet the minimal requirements for stakeholders performing modeling tasks.[10]shows that the typical shortcomings of traditional BPM approaches can be overcome with S-BPM.The concept of S-BPM to partition a process into subjects leads to a clear separation of concerns from a stakeholder point of view and also to well-defined communication interfaces based on incoming and outgoing messages,which both foster the possibility for reuse in new models.Therefore,it would be desirable to choose SBPM notations as our model representation to solve the spatial variability issue,which will make our solution simplified and natural.

In this paper,we propose an organization-driven business process configurable modeling approach.This idea follows Conway’s law[11]and is especially suitable for the organization-dependent context.In the proposed configurable framework,activities are associated with organizations and data information.The configuration starts with selection of organization departments(i.e.,subjects),then tasks,and at last the data information.In summary,this work has the following two main contributions:

-An organization-driven business process configurable modeling approach is proposed for spatial variability by integrating both restriction and extension operations based on the S-BPM paradigm.By applying our approach,a new process model variant with specific data models can be derived to adapt to a new application scenario.

-We substantiate the practicability of our approach using a real generic example,which demonstrates the effectiveness of the approach.Our work contributes to promoting the reuse of both process models and services,thus relieving the development task,and making process modeling easier and efficient.

The rest of the paper is organized as follows.We provide a motivating example in Section II.Section III presents the fundamentals of S-BPM.In section IV,we propose our organization-driven process configurable modeling approach in detail,including the configurable S-BPM model,and its configuration steps accompanying a running example to demonstrate the effectiveness.Section V gives a discussion about the proposed approach.Section VI reviews the related work on configurable modeling for business processes.Finally,the paper concludes with a summary and presents some research issues for future work in section VII.

II.MOTIVATING EXAMPLE

To motivate the pressing need of process modeling for spatial variability,we present a simplified example in this section from a real business process with many variants developed for the Real Estate Administration(REA)in a certain province of China.At present,REA is attempting to promote its business in many other cities of China.

The example process is to apply public houses for a special group of people,e.g.,those who have low income or suffer natural disasters.After an application form is accepted,it will go through several review stages or investigations by different governmental departments from a low level to a high level.During the process,some data information may be needed by investigations.Because of the different customer needs,government regulation,and organization structures in different cities,different process model variants of the business are generated,thus hindering the market promotion of the process information system.In Figure 1,(a),(b),and(c)are three process variants of the public housing application(PHA)process in BPMN for three cities:cityA,cityB,and cityC(Note that for the reason of confidentiality,we do not use the real city names in this paper).The organization’s information is marked by background colour on activities.That is,the activities with the same background colour are performed by the same organization department.The governmental organizational structure is decentralized,while the process-aware information system can be both centralized or decentralized.

As can be seen from figures(a),(b),and(c),they have different sets of activities for the same business.Based on the understanding of the business in three cities,we know that the processes have 5 common activities:Accept App-Form,First review,Secondary review,Final review,and Archive.These activities for a business are relatively fixed even though they may be named differently and associated with different organizations in different cities.While other activities are relatively variable in different cities.Such variability comes from the difference in organization structures.One organization exists in a city,while it may not exist in another city.And sometimes organizations may merge or split due to the change of government policy in one place.In general,different organizations have different authorization in a city,and one organization in different cities may also have different authorization or capability.Therefore,different organizations(or delegations)lead to different sets of activities.Under such circumstances,different municipalities need to offer the same set of services to their citizens and support many variants of the same process.

However,the existing works on configurable process modeling lead to complex configurable process models when the number of model variants increases or when activities are differentiated by their names.Thus,designers are in a dilemma to do the configuration for the desired result due to a lack of organization information.It would be worse in organization dependent context since organization structure is closely related to process tasks.The number of tasks and sequences in different cities’scenarios may be different.As a result,the complete process model for a new scenario is hardly derived from the corresponding configurable process model due to a lack of organization information.Based on these observations,we argue that it is preferable to focus on the causes rather than the results.Thus,it is pragmatic to specify organization information in the configurable process model.In our approach,organizations are considered first,thereby the tasks and data models are determined further.We will see a configurable process modeling procedure for the PHA process in section IV,by which the process for a new city can be generated through configuration.

III.S-BPM FUNDAMENTALS

S-BPM focuses on the acting elements of a business process in contrast to traditional BPM.The acting elements are called Subjects.They encapsulate their behaviour in the form of a local process model.S-BPM provides a framework that focuses on the subjects of a process.The S-BPM notation is based on natural language with complete sentences comprising subject,predicate,and object,and can be directly executed.All people involved in the processes of an organization can learn how to model subject-oriented process models.Modeling in S-BPM takes place in two stages:

Figure 1.Public housing application processes in cityA,cityB,and cityC.

⋄identification of the subjects and their interactions→What are the subjects?

⋄Definition of the subject behaviour→What does a subject do,and when does it send or receive a message?

In the first step,the subject-oriented description of a business process is composed of subjects and messages,along with required data that are exchanged between the roles involved.For example,the interaction between subjects in a business trip process model is illustrated in Figure 2,which is calledSubject Interaction Diagram(SID)[9].SID is exactly the result of the first step.

After describing the interaction between subjects,the behaviour(including activities and interactions)of individual subjects is described.In the second step,the subjects are refined by modeling their behaviour as a sequence of activities and interactions with states and transitions.Several subjects can act in parallel and synchronize themselves through messages.The behaviour of every subject can be specified using three states(i.e.,Send,Receive,and Function)and transitions.Function state indicates an action,such as fill out Bt-request,of the subject.Send and Receive states indicate interactions with other subjects via the exchange of messages.The execution path is defined via transitions between the modeled states in the subject behaviour.S-BPM models are represented graphically using only five symbols,namely the three state types,messages,and subjects[12].

Figure 2.Interaction between subjects in the application process.

For example,the description of behaviour for the employee in a business trip process model is shown in Figure 3,which is calledSubject Behaviour Diagram(SBD)[9].The SBD of a subject is a Finite State Machine(FSM)graph,describing the sequence of executing actions,sending and receiving messages.Therefore,S-BPM models are represented graphically using five symbols,namely the three state types,messages,and subjects[12]as illustrated in Figure 2 and Figure 3.It was proved[9]that each syntactically correct S-BPM model can be interpreted by an appropriate work flow engine.Thus,the graphical notation is based on a clear formal semantic,allowing for automated code generation(executable models)[10].

Figure 3.Illustration of the employee’s behaviour in business trip process.

IV.ORGANIZATION-DRIVEN BUSINESS PROCESS CONFIGURABLE MODELING

In this section,we will propose the configurable SBPM model first.The main idea of our approach is that the configurable S-BPM process model generate by merging S-BPM process model variants,and modeling new S-BPM process models by configuration from the the configurable S-BPM process model with both restriction and extension operation.The specific configuration steps driven by organization structure are introduced with a running example throughout this section for demonstration.

4.1 Configurable S-BPM Process Model

Borrowing from the idea of the existing configurable model approach for traditional activity-centric process models,i.e.,variability by restriction,we try to merge the S-BPM model variants to be an integrated configurable model.before that,we identify the distinct subjects,and functions coming from the existing S-BPM process model variants for the target business.

The activities in BPMN models correspond to functions in S-BPM models.An activity may be named differently in different process variants,thus we unify the activity names by comparing their similarity to match graph elements for later merging.Let a∈A1be an activity in the process variant P1,where A1is the set of activities in P1.Likewise,a′∈ A2be an activity in the process variant P2.To compute the similarity SimAbetween a and a′,we use a combination of syntactic and semantic similarity metrics since they are popular for measuring the similarity between activity labels in process models[13].We use a syntactic similarity based on Levenshtein distance[14]which computes the number of edit operations(i.e.,insert,delete,or substitute a character)needed to transform one string into another.For semantic similarity,we use the WordNet database[15]which is a lexical database for English words.The WordNet similarity package includes a set of algorithms for returning the synonyms between two words.We use in particular the WUP algorithm[16]which measures the relatedness of two words by considering their depths in the WordNet database.After normalizing activity labels(i.e.,put all characters in lowercase,remove stop words,etc.),the total similarity is the average of their syntactic and semantic similarities.

After computing the similarity between activities from different process variants with the Eq.(1),we unify similar activities with the unified names manually.As for the three process variants as illustrated in Figure 1,even they are described in BPMN,the nameunified functions(i.e.,activities)are presented in Table 1.There are 12 unique functions in total after unified naming.Similarly,two subjects may be named differently in different process variants but they own the same or similar function set.In such a case,we use the Jaccard Coefficient[17]to compute similarity over two subjects,as shown in the Eq.(2).

where FSAand FSBindicate the function sets of subject SAand SBrespectively.As for the process variants illustrated in Figure 1,we get the following 6 subjects:S1–Working Unit,S2–Community,S3–Street,S4–County,S5–District Office,S6–Municipal Bureau,if they are differentiated only by names.A distinct function could be done by different subjects in different process variants sometimes.As for the example in the PHA process,we derive the subjects and their functions as shown in Figure 4,by which we could compute the similarity between two subjects if they share common functions.Consequently,we get SimS(S2,S6)=1/6 and SimS(S4,S5)=1 by Eq.(2).If we take two similar subjects as one when their similarity is larger than 0.8,then it is inferred that County and District Office are in fact one subject.Thus,we get 5 distinct subjects finally.

With the name-unified functions,we can derive the corresponding S-BPM models for the three process variants in Figure 1.We present an integrated view for S-BPM models for cityA,cityB,and cityCrespectively as illustrated in Figure 5,6,and 7,in which SID and SBD views are combined.

Then with the S-BPM process variants,we merge their SID view and SBD view for each distinct subject to get the configurable S-BPM model.In essence,SID and SBD are both directed graphs.The merger of two directed graphs can be achieved by union operation of sets of vertexes and edges. We define merger operation for directed graphs as Def.1.

The merging result in Def.1 is also a directed graph including all vertexes and edges from the two original directed graphs.Since the union operation of sets meets commutative and associative laws,thus the merger operation for directed graphs is also commutative and associative,i.e.,the following equations hold.

Thus,＜G,⊕ ＞ is an algebra system which meets the closeness,where G is a set of directed graphs representing SIDs or SBDs.Therefore,the merging result of n directed graphs is irrelevant with the operation sequence.When there is a new process model variant,it can be merged into the obtained result later.As for the running example,the merged SID view is shown in Figure 8,which is also the resulting configurable SID.And the resulting merged SBD views for the distinct subjects are illustrated in Figure 9 from process variants.

Each function may operate on some business objects.The mapping between the business objects and the name-unified functions for the PHA process can be seen from Figure 10.specifically,functions Accept App-Form,Conducting survey,Cooperative investigation map to business objects Application Form,Questionnaire Form,Investigation Form respectively.When one of them has been done,an associated business object may be created.In this running example,functions Public notification,Archive do not write on any business object while events will be generated after they have been done.And the rest of the nameunified functions write to a data field of Application Form respectively.Note that other default or necessary data fields may be written when the associated business objects are created.In summary,the resulting complete configurable S-BPM model consists of three parts:the merged SID view,the merged SBD views,and the mapping model as well.

Table 1.Collected name-unified functions for the PHA process.

Figure 4.Subjects and their acting functions.

Figure 5.PHA process in S-BPM approach for cityA.

4.2 Modeling of S-BPM Process Model by Configuration

Driven by the idea of“Design by Reuse”,we use the configurable S-BPM model to build S-BPM models for a new scenario.Suppose that REA plans to deploy the PHA process in a new city like cityDfor example.The target process model to be deployed in BPMN is shown in Figure 11.Next,we will present how to obtain the target S-BPM process model by configuration operations from the configurable S-BPM model.

As mentioned before that a S-BPM model consists of two types of representations:a SID and a set of SBDs.The SID includes the subjects,the messages exchanged between the subjects and the business objects attached to the messages.The SBD includes all possible state sequences of a subject:a finite set of Send,Receive,and Function states.Next,we will model the SID,SBDs,and business objects step by step.

4.2.1 Modeling SID

In the first step,a specific organization structure will be the input,so that the modeler will know which organization departments exist and which ones should be included in a concrete business scenario.At this stage,a local official in charge of the business is needed to provide domain knowledge about which departments should be involved to deal with the business process.Because the definition of which subjects should be part of a process is a leadership decision–this is why the Governors(i.e.,subjects who have responsibility for environmental factors and who take influence on the respective work and development processes)need to be involved[9],while the guidelines often stem from the organization structure.

After the involved subjects are determined,both restriction and extension approaches are used for modeling SID starting with the merged SID from process variants.The procedure may require several steps:

Figure 6.PHA process in S-BPM approach for cityB.

Figure 7.PHA process in S-BPM approach for cityC.

Figure 8.Merged SID view.

a)Replace a similar subject by another name with reference to its function set(restriction);

b)Remove subjects which are not necessary(restriction);

c)Remove message connections between subjects which are not necessary(restriction);

d)Add required subjects which are not included in the merged SID(extension);

e)Add required message connections between subjects which are not included(extension).

Note that,removing/adding the message above includes removing/adding the corresponding attached business object.From the above steps,SID may be derived not only by restriction but also by extension sometimes.While most of the time,we believe only restriction operations are enough.As for the target scenario in cityD,its required organization departments for the business are Working Unit,Community,Street,County,Municipal Bureau.Thus,these five organizations are selected as the corresponding five subjects.The con figured SID derived by restriction for cityDis shown in Figure 12,where the removed connection is indicated by a gray solid line with an arrow.The business object exchanged between subjects is the Application Form,so we omit the business objects to attach to the messages.

4.2.2 Modeling SBDs

After modeling SID,it is known that how the involved subjects should interact with each other.With the merged SBDs,the modeler could model SBD for each involved subject also by both restriction and extension approaches.S-BPM enables the subjects to create their SBDs respectively,within the guidelines specified by the Governor,themselves,and thus to actively design the development of the respective processes[9].This configuration way by restriction and extension makes modeling SBDs easy.Because when a new business scenario comes,subjects may not be familiar with it much.

Functions are selected from its merged SBD for each subject according to the authorization of the organization department,and receive and send states as well.Note that one function will be included only in at most one resulting SBD,even though one function may be included in different merged SBDs sometimes.The procedure may require several steps:

a)Remove states(Function,Receive,and Send)which are not necessary(restriction);

b)Remove transitions between states which are not necessary(restriction);

c)Add required states(Function,Receive,and Send)which are not included in the merged SID(extension);

d)Add required transitions between states which are not included(extension).

At this stage,if an additional function is needed,then it could be added directly to model the subject behaviour.And if there is any added subject or function when modeling SID or SBDs,it is suggested that the resulting S-BPM model should be merged into the configurable model.In this way,the configurable SBPM model is maintained with time.SBDs for the required subjects are illustrated in Figure 13.The dashed lines indicate the synchronizations between subject behaviours,which are also shown in Figure 12.By comparison,we could know that the S-BPM process model for cityDin Figure 13 is equivalent to the process model from Figure 11 to achieve the same business goal.

4.2.3 Modeling Business Objects

In this example of the PHA process,when the S-BPM process model is derived,the business object Application Form is decided based on the dependency relationship from Figure 10.Except for the basic infor-mation of an applicant that is submitted to the community,other data fields are needed to store the review and evaluation results.While the business object Questionnaire Form is relatively fixed since it is not written by any other functions after it is created.And if there is a reference template for the business object,we can edit it quickly based on the required functions with write operation.The con figured business objects of the PHA process for cityDare shown in Figure 14.

Figure 9.Merged SBD view for subjects Community,Working Unit,Street,County,and Municipal Bureau.

Figure 10.The mapping between business objects and functions for the PHA process.

Figure 11.Target PHA process model in BPMN for cityD.

As subjects get more familiar with this business,they can adapt their dynamic behaviour further in the context of an evolutionary business information system.Since control flow is always local to a subject defined in its internal behaviour,changes can be performed without affecting other stakeholders(as long as these changes require no adaptation to exchanged messages).Thus,S-BPM is more suitable for modeling variable business process models in both design time and run time than traditional BPM.

Figure 12.Derived SID for cityD.

V.DISCUSSION

From the running example presented along with the introduction of our approach,the aim for business process modeling for spatial variability is achieved.Based on the S-BPM approach,the process modeling becomes simplified and concerned in a subject.Therefore,the effectiveness and usefulness of our approach are validated from the perspective of process modeling.If we take functions as services from the perspective of SOA(Service-Oriented Architecture)[18],all the required services for cityDcan be reused from process variants,thus the development cost can be relieved.In fact,only functions added by extension operations would be developed in our approach,while this situation rarely occurs in the configuration in general cases.Therefore,from the perspective of process system development,our approach has better efficiency compared with the traditional activity-centric process modeling.

Although the configurable S-BPM model could help designers model S-BPM processes easily and efficiently,and the post-processing work after a configuration operation can be done automatically,the configuration selection is still a manual procedure according to the personalized requirements.Thus,some guideline from domain knowledge is needed,especially for complex application scenarios.Machine learning or data mining techniques may be considered to mine some potential domain knowledge or configuration rules.From another perspective,if the personalized requirements can be transformed into the formal input,the configuration procedure may be done automatically.Besides,quantitative evaluation for the proposed approach is needed from practical applications.

VI.RELATED WORK

In practice,business process models are often reused and evolved in different application contexts,resulting in a number of related process model variants.Typically,these process variants pursue the same or similar business goal and share several commonalities.However,they can also exhibit some variable properties,depending on how they are used in different application contexts.Thus,process modeling for variability becomes necessary.Rich works on this topic have been done.By reviewing these works,we divide them them roughly into two classes:variability by restriction,andvariability by extension.In the following,we will detail them one by one.

Variability by Restriction:Variability by restriction starts with a customizable process model that contains all behavior of all process variants.Customization is achieved by restricting the behavior of the customizable process model[2,19].In the past,one may model process models from a reference process model.However,reference process models in commercial use lack a specification of variation points and configuration decisions.This shortcoming is addressed by the concept of configurable process model[3]for various activity-centric process modeling languages,such as C-SAP WebFlow,C-BPEL,C-YAWL,C-EPC[20].This concept is a step forward towards the systematic reuse of(reference)process models.Typically,as EPCs are widely used for reference process modeling[21],an extended reference modeling language,named configurable EPCs(C-EPCs),is proposed by M.Rosemann et al.[3].Further,M.La Rosa et al.extends the original EPC notations by associating roles and objects to functions,in an integrated EPC(iEPC),and the corresponding configurable iEPC(C-iEPC)is also proposed[21].Alternatively,obtaining a configurable process model can be done via the merger of process model variants. Some techniques are proposed in literatures.Gottschalk[20]elaborates on the merger of process models into a single configurable process model.Li et al.[22]present an approach for creating a new reference model based on models mined from an event log.Mendling et al.[23]present an approach to merge the different views on process models.Sun et al.[24]focus on merging block-structured process models.These techniques allow for the instantiation of an unsound process model from a configurable process model.Furthermore,some techniques are not reversible.Thus,Dennis M.M.et al.[25]capture the control- flow of a process by a CoSeNet:a configurable,tree-like specification of the process model,which is sound by construction.Recently,Wehbi et al.[26]propose a knowledge-based approach for capturing process model versions into a single configurable business process and splitting this process into fragments is also discussed.The approach builds a configuration knowledge base to track the business process variability(ie,particularities of each business process variant).Variants are represented as a new configurable process structure tree resulting from fragmenting a business process.

Figure 13.Derived SBDs for cityD.

Figure 14.Con figured business objects of PHA process for cityD.

Variability by Extension:Variability by extension takes the opposite starting point.The customizable process model does not contain all possible behavior,instead it represents the most common behavior or the behavior that is shared by most process variants[2].Different from configurable modeling by merge,Alena et al.[7]state that particular process variant can be con figured at a high level of abstraction by applying a set of well-defined change operations to a reference process model and a Provop(process variants by options)framework is proposed for enabling automated configuration of process variants.Similarly,Riccardo et al.[8]propose to express variability in business process models by extending the feature model from software product line.Besides,there are some works on process modelling for variability,such as[27,28],while they focus on artifact-centric modeling paradigm[29,30].

In a nutshell,the existing configurable approaches are either variability by restriction or by extension.However,the configurable model may become large and complex due to the number of process variants,thus it is hard to use for modelers.Moreover,even the de facto standard BPMN is not easy to be understood by people who are working in a process[31].Last but not the least,these approaches neglect the organization structure’s impact on process modeling,which is an especially significant factor and is the motivation in our research scenario.In the PHA process scenario,the organizations can decide which tasks have to be done,and each organization will decide the sequence of its involved tasks.Therefore,an easy to use and organization-driven process modeling approach is needed.

VII.CONCLUSION

In this paper,we propose a configurable business process modeling approach for spatial variability driven by organization structure.The organization departments correspond to subjects in the S-BPM modeling paradigm,thus it fits organization dependent context well.First,we merge S-BPM process variants to obtain the integrated configurable S-BPM model,including the merged SID view,the merged SBD views,and function-data mapping model.Further,configuration steps are presented in detail with a running example to demonstrate the effectiveness.The con figured results are SID,SBDs,and business objects.The resulting SBPM model allows for immediate execution(testing)with S-BPM tools.Our work contributes to promoting the reuse of both process models and services,thus relieving the development task,making process modeling easier and more efficient for new application environments.

Our configurable S-BPM business process modeling approach for spatial variability is founded on organization dependent context.More specific selection guidelines or constraints of functions for subjects may be needed when it comes to a new business at the beginning.In addition,our configurable process modeling approach focus on modeling variability.We will study the formal analysis and verification for S-BPM process models in future work.

ACKNOWLEDGEMENT

The authors appreciate the anonymous reviewers for their extensive and informative comments for the improvement of this manuscript.This work was supported by National Key R & D Program of China under grant No:2020YFB1707602,Educational Commission of Hunan Province of China under Grant No:20B244,National Natural Science Foundation of China under grant No:61872139.