Generating Persistence Structures for the Integration of Data and

Control Aspects in Business Process Monitoring

Eladio Domínguez

, Beatriz Pérez

, Ángel L. Rubio

, María A. Zapata

Alberto Allué

and Antonio López

Departamento de Informática e Ingeniería de Sistemas, Universidad de Zaragoza, Zaragoza, Spain

Departamento de Matemáticas y Computación, Universidad de La Rioja, La Rioja, Spain

Infozara Consultoría Informática, Zaragoza, Spain

Keywords:

Model–driven Approach, Development of Monitoring Systems, Process Flow and Data Integration, UML,

Occurrence–centric.

Abstract:

Today’s organizations have to monitor increasingly complex business processes that handle large amounts

of data. In this context, it is essential to design working frameworks that seamlessly integrate both control

ﬂow and data perspectives. Such an integration can be eased by automatically generating the infrastructures

for storing data and control aspects. Towards this goal, we propose an automatic process for synthesizing

persistence structures for control ﬂow and data storage. In particular, based on an approach centered on

the concept of Occurrence, in this paper we present a proposal by means of which, after applying several

translation patterns to a business process model, we automatically generate the persistence structures that

integrate both data and control aspects of such model. The feasibility of this proposal is demonstrated by

developing a prototype and evaluating its application to different examples taken from the literature as a

benchmark.

1 INTRODUCTION

Continuous improvement is an integral part of the life

cycle of any business process. In order to achieve this

objective, monitoring systems of process execution

are required. From the information obtained through

monitoring, it is possible to assess the adequacy of

the processes to the reality and to proceed to the re-

vision of the underlying models. Due to its intrinsic

nature, business process management systems must

be able to capture both the control ﬂow perspective

and the data perspective in order to faithfully achieve

their goal. This is even more important when mo-

nitoring tasks are performed, in which the managed

data must be aligned with the enacted processes. Of-

ten, these two perspectives have been treated separa-

tely, but more and more it is recognized (Eshuis and

Van Gorp, 2016a; Eshuis and Van Gorp, 2016b; Ku-

maran et al., 2008; Künzle and Reichert, 2011; Ni-

gam and Caswell, 2003) that a holistic and as much

automated as possible approach of both aspects is re-

quired.

In (Domínguez et al., 2017; Domínguez et al.,

2014) we presented an approach based on the no-

tion of Occurrence, a three–dimensional modeling ar-

tifact that comprises guidance, structure and behavior

(see Section 2). The ﬁnal goal of that proposal is to

improve the processes of design and development of

monitoring systems, so that analysts and developers’

work is lightened.

In this paper we propose a further step towards

the development of monitoring systems that seamles-

sly integrate control ﬂow and data aspects. We deve-

lop an improvement of the design strategy presented

in (Domínguez et al., 2014), enriching it with auto-

matic mechanisms for the construction of storage in-

frastructures for monitoring issues. This proposal has

several potential advantages such as minimizing co-

ding tasks and simplifying the management of the life

cycle of business processes. In particular, in Section 3

we show how from a business process model, re-

presented through a UML activity diagram that con-

tains information about business objects (other busi-

ness process modeling languages could also be used),

a speciﬁc class model can be automatically generated.

This class model, created by applying a UML proﬁle

and several patterns, comprehensively captures ﬂow

and data perspectives. Section 4 shows the details of

320

Domínguez, E., Pérez, B., Rubio, Á., Zapata, M., Allué, A. and López, A.

Generating Persistence Structures for the Integration of Data and Control Aspects in Business Process Monitoring.

DOI: 10.5220/0006781903200327

In Proceedings of the 20th International Conference on Enterprise Information Systems (ICEIS 2018), pages 320-327

ISBN: 978-989-758-298-1

ContextClass_C1

ContextClassC1_State

ContextClassC2_State

0..1 0..1

1 1

0..1

ProtocolExecutionP1

ProtocolExecutionP2

ActionA22

ActionA21

Action_A11

Action_A12

0..1 0..1

0..1

* *

ContextClass_C2

ProtocolExecution

«ProtocolExecution»

ContextClass

ContextClassState

actual

0..1

initial

«ContextClass»

ConcreteState_S11

«StateInstance»

ConcreteState_S12

«StateInstance»

ConcreteState_S21

«StateInstance»

ConcreteState_S22

«StateInstance»

«TransitionInstance»

ConcreteTransition_S11_2_S12

{disjoint, incomplete}

source

target

«TransitionInstance»

ConcreteTransition_S21_2_S22

Figure 1: OcBase PIM pattern.

the translation and the automatization procedures of

our approach, together with an implementation into

an MDE–based tool, using the Atlas Transformation

Language (ATL) (ATL Plug-in, 2017).

The proposal has been evaluated according to se-

veral criteria, such as performance and maintenance

(Section 5). Other related works are discussed in

Section 6, while Section 7 concludes with ﬁnal re-

marks and future work.

2 BACKGROUND: THE

OCCURRENCE– BASED

APPROACH

Business process monitoring is concerned with recor-

ding information about the actual execution of pro-

cesses in order to later extract valuable knowledge

that can be utilized for business process quality im-

provement (Campanile et al., 2008; Kang et al., 2012;

Kang et al., 2009; Reichert et al., 2010). The type

of collected information inﬂuences the quality of the

diagnosis accomplished during the monitoring phase.

Trying to improve the diagnosis, in (Domínguez et al.,

2017; Domínguez et al., 2014) a holistic perspective

of system dynamics is proposed which enhances the

monitoring of business processes as a whole. Spe-

ciﬁcally, when the dynamics of a system is deﬁned

as a set of processes accomplished by following se-

veral protocols, we suggested a three–dimensional

business process monitoring approach, which simul-

taneously encompasses structure (objects), behavior

(events, states, state changes) and guidance (proto-

cols, processes and protocol performers).

According to this perspective, each protocol,

when executed, acts on one or more objects, which

react by changing their state and/or related data. An

Occurrence is a concrete, identiﬁable and indivisi-

ble chunk of information which contains aspects or-

ganized according to three dimensions: guidance,

structure, and behavior, given by protocol execution,

object, and effect, respectively (Domínguez et al.,

2014). The proposal also deﬁnes persistence struc-

tures that simplify the subsequent processing of the

system trace. Occurrences, generated through proto-

col executions, are stored under a speciﬁc persistence

structure, called Occurrence Base (OcBase). Besi-

des, an Occurrence Management System (OcSystem)

is any system that enables the comprehensive mana-

gement of occurrences, both in terms of their storage

by using an OcBase, and in terms of their handling

by using embedded tools that ease the exploitation of

derived knowledge.

This proposal provides several theoretical con-

cepts that can be applied using different methodologi-

cal and/or technological approaches. In (Domínguez

et al., 2014) we proposed a speciﬁc design strategy

for the development of an OcSystem following an

MDE approach (called OcBaseMD design strategy).

In particular, this strategy provides several artifacts

(the Occurrence proﬁle and four patterns) for the de-

sign of an Ocbase, that is, for the design of the persis-

tence structures that will store the occurrences genera-

ted through system execution. The main pattern pro-

vided by this strategy is the OcBase platform indepen-

dent model (PIM) pattern (see Figure 1). This pattern

makes use of the Occurrence proﬁle which consists of

four stereotypes corresponding to the executed proto-

cols («ProtocolExecution»), acting on objects («Con-

textClass»), which react by changing their state («Sta-

teInstance» and «TransitionInstance»). Furthermore,

this pattern establishes a skeleton of the OcBase class

diagram in order to capture the hierarchical struc-

ture of state classes, and the transition classes among

source and target state classes. Several advantages are

obtained from using this design strategy that make

it different from other monitoring proposals (Cam-

panile et al., 2008; Kang et al., 2012; Kang et al.,

2009; Reichert et al., 2010). Three of the most im-

portant advantages are: (1) a holistic perspective of

system dynamics is provided, enhancing the monito-

Generating Persistence Structures for the Integration of Data and Control Aspects in Business Process Monitoring

321

p1: Patient

[NotObese]

Calculate Body

Mass Index

Give lifestyle

suggestions

p1: Patient

[ChangingLifestyle

AndBehaviour]

<<Monitored>>

Discuss drug

risks

Don't give

drug treatment

Apply drug

therapy

p1: Patient

[WithoutDrug]

p1: Patient

[WithDrug]

Assess patient

conditions

Discuss surgery

risks

Don't give

surgery

Plan

surgery

p1: Patient

[PreparedForSurgery]

p1: Patient

[WithoutSurgery]

Perform

surgery

p1: Patient

[WithSurgery]

<<Monitored>>

p1: Patient

[Obese]

p1: Patient

[Discussing

DrugRisks]

p1: Patient

[Discussing

SurgeryRisks]

<<Monitored>>

Assess comorbid

conditions

Doctor

visit

[<30]

[>=30]

Assess patient

weight

<<Monitored>>

Figure 2: UML Activity Diagram of the obesity clinical guideline.

ring of business process as a whole (Domínguez et al.,

2014), (2) an occurrence query framework is available

which simpliﬁes the querying process for obtaining

provenance-related information (Domínguez et al.,

2017), and (3) a single–pool solution for OcBase evo-

lution is proposed for tackling the highly probable

changes in the processes without incurring any lost

of existing data (Domínguez et al., 2014).

In our research and development group (incorpo-

rating partners from both academia and industry), this

approach has been applied to develop successful mo-

nitoring applications, such as a biobank OcSystem

for identifying genetic and lifestyle aspects related to

cardiovascular risk factors (Domínguez et al., 2017;

Domínguez et al., 2014). Another example is the QRP

Platform which is a Capability Maturity Model Inte-

gration (CMMI) appraisal tool for project quality ma-

nagement (Allué et al., 2013).

3 SYNTHESIZING PERSISTENCE

STRUCTURES FOR CONTROL

FLOW AND DATA STORAGE

We consider that, within the framework of a design

strategy, the proposals for developing monitoring sy-

stems must include an automatic process for transla-

ting business process models into persistence structu-

res. In this paper, we enrich the OcBaseMD design

strategy presented in the previous section by provi-

ding it with a translation process that, starting from

a process model, generates the storage infrastructure

of an OcBase for storing the occurrences generated

during the execution of that model.

Business process modeling can be performed

using different conceptual modeling languages, such

as BPMN (Object Management Group, 2011) or

UML Activity Diagrams (Object Management Group,

2015), among others. The application of the previ-

ously presented occurrence–based approach requires

a language that facilitates the representation not only

of the control ﬂow but also of the objects (structure)

and their state changes (behavior) produced during

the process execution. UML Activity Diagrams ve-

rify this requirement since it is a conceptual business

process modeling language that allows us to specify

the ordering of activities (control ﬂow) as well as the

ﬂow of objects (data ﬂow) that are used by the activi-

ties. In this paper, as an example, we assume that the

business processes are represented by means of UML

Activity Diagrams, but other languages would also be

possible.

The control and data ﬂows represented in a UML

activity diagram are used for determining the type of

elements that will be stored in the OcBase for mo-

nitoring purposes. With the goal of allowing the de-

signer to take part in this decision, we have deﬁned

the Monitoring proﬁle. This proﬁle consists of a ste-

reotype by means of which the designer, bearing in

mind the monitoring requirements, can determine the

objects and states that must be processed. The ste-

reotype is called «Monitored» and extends the UML

ObjectNode element. In this way, only the object no-

des of the UML activity diagram with the «Monito-

red» stereotype will be taken into account during the

generation of the OcBase schema. For example, Fi-

gure 2 shows a stereotyped UML activity diagram re-

presenting a real–life clinical guideline for the mana-

gement of obesity in primary care (Snow et al., 2005).

According to this guideline, different actions can be

performed when a patient has a body mass index of

30 Kg/m

or greater. Note that several object nodes

of type Patient appear in this activity diagram, but

only some of them are monitored. We use this guide-

line as case study along the paper.

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

322

...

Object

[A]

Object

[B]

...

Monitored

StateInstance

TransitionInstance

A2B

transition pattern

Object

[A]

Monitored

state pattern

StateInstance

Object

[A]

«Monitored»

...

Object

[B]

Monitored

Object

[C]

Monitored

Object

[D]

Monitored

AND state pattern

«StateInstance»

StateInstance

«StateInstance»

region 1

region 2

ANDstate 1

Figure 3: Translation patterns.

Stereotyped activity

diagram model

Stereotyped class

diagram model

Occurrence Profile

Model

Transformation

Figure 4: Translation process scheme.

The method for automatically generating the

storage infrastructure of an OcBase (see Figure 4)

takes as source models an activity diagram stereoty-

ped with the Monitoring proﬁle, and the Occurrence

proﬁle that will be applied to the target class dia-

gram. Starting from these two models, the transla-

tion process synthesizes the UML class diagram of

the OcBase, replicating the structure of the PIM pat-

tern (see Figure 1).

The translation process takes the structure of the

PIM pattern as scheme, so that several elements of

the generated UML class diagram are common to

all OcBases, independently of the designed UML

activity diagram. Examples of these elements are

the ContextClass, the ContextClassState and the

ProtocolExecution classes. On the contrary, ot-

her elements of the OcBase, mainly the hierarchical

structure of state classes, and the transition classes

among source and target state classes, are determined

depending on the given UML activity diagram. These

activity diagram’s dependent elements are created ap-

plying several translation patterns. Next, we explain

three of the most outstanding patterns (see Figure 3).

State Pattern. For each monitored object node of

the activity diagram, with state A, a class A represen-

ting the state is created in the OcBase.

AND State Pattern. This pattern is applied when

there are stereotyped object nodes of the same ob-

ject, entering or exiting activities which form part

of a fork/join structure. In this case, a hierarchical

structure of classes with one abstract class, and a re-

gion class for each ﬂow of the fork, is created in the

OcBase.

Transition Pattern. This pattern is applied when

there is a pair of stereotyped object nodes of the same

object in different states (A and B), one of them en-

tering an activity and the other one exiting another

activity (with no other object node of the same object

in the intermediate activities). A class A2B represen-

ting the transition between the states is created in the

OcBase.

4 AUTOMATIZATION OF THE

TRANSLATION

In order to demonstrate the feasibility of our transla-

tion proposal, we have implemented it into an MDE–

based tool, using the ATL Eclipse plug–in (ATL

Plug-in, 2017), developing a ﬁrst prototype. Speciﬁ-

cally, the translation process has been implemented as

an ATL module called ActivityD2ClassD, together

with a simple ATL library called UtilityLibrary.

The source models of these ATL units (that is, the ste-

reotyped activity diagram and the Occurrence proﬁle)

must be provided in XMI syntax as .uml extension

ﬁles and have to conform to the UML 2.5 metamo-

del. By using the deﬁned units, the activity diagram

is translated into an OcBase class diagram in XMI

syntax, also conforming to the UML 2.5 metamodel.

A complete explanation of our implementation of the

translation process can be found in (Supplementary

Material, 2018).

While the activity diagram could have monitored

object nodes of any type, our prototype focuses on ob-

ject nodes that refer to the same object (the translation

of other design alternatives are discussed later in this

section). Based on this, before performing the transla-

tion, the prototype must be provided with the informa-

tion regarding the object and its corresponding type.

Since ATL does not supply a way to parameterize va-

riables’ values at runtime, we include the correspon-

ding pair of data object-type into an ATL helper deﬁ-

Generating Persistence Structures for the Integration of Data and Control Aspects in Business Process Monitoring

323

Figure 5: Generated Class diagram for the obesity case study.

ned within the independent UtilityLibrary library.

In this way, not only do we separate such required

information from the remainder translation, but also

we anticipate to possible future improvements of the

algorithm.

The ATL ActivityD2ClassD module, on the ot-

her hand, constitutes the pivotal part of the trans-

lation, implementing our patterns and using the in-

formation included in the UtilityLibrary library.

The module consists of a set of 39 ATL helpers,

which act as global variables and auxiliary methods,

1 matched rule and 17 called rules (around 1000 to-

tal lines of code). The strategy followed herein for

implementing our translation patterns relies mainly

on two of these rules: the createPackage and the

traceFromInitialNode.

The called rule traceFromInitialNode consti-

tutes the heart of the translation. The main aspect

of this rule is that it is deﬁned in a recursive man-

ner, in such a way that in each call a new element

in the source diagram is tackled. This rule performs

two main tasks. Firstly, it traverses the overall acti-

vity diagram going through both the data and control

ﬂow, identifying the situations where our patterns can

be applied. Secondly, as it traverses the activity di-

agram, it gathers and determines the information re-

garding the new elements in the class diagram to be

created, which are activity diagram’s dependent.

The second rule called createPackage is the

only matched rule and starts the translation pro-

cess. This rule is in charge of generating the over-

all stereotyped class diagram, including both the ele-

ments which are common to all OcBase class di-

agrams and those which are activity diagram’s de-

pendent elements. The ﬁrst type of elements (that

is ContextClass, the ContexClassState and the

ProtocolExecution classes) are directly generated

and the second type of elements are generated using

the traceFromInitialNode rule.

The application of our prototype to the activity di-

agram of Figure 2 results in the stereotyped class dia-

gram of Figure 5, which replicates the PIM pattern

including, for example, the Patient context class,

the WithSurgery state class or the transition class

DiscussingSurgeryRisks2WithSurgery.

With regard to the scope of our proposed pro-

totype, as described previously, it focuses on nodes

that refer to the same object of a speciﬁc type. The

prototype could be easily extended to deal with other

design alternatives such as activity diagrams with dif-

ferent object nodes referencing objects o

, o

, etc. of

a same object type t, or activity diagrams with mul-

tiple object types (t

, t

, etc.). In both situations the

main issue corresponds to establishing the different

hierarchy and source-target structures deﬁned by the

states of the different object nodes. In the ﬁrst situa-

tion, we just need to register independently, per each

object o

, the structures deﬁned by its object nodes

through their different states. The second situation

could be implemented similarly since the object no-

des of a speciﬁc object type t

would lead to indepen-

dent parent-child and source-target structures. An im-

plementation alternative could be to trace the diagram

once per each object type and later join the results.

As for Activity diagrams’ elements supported by

our prototype, we would like to note that currently

it does not support the existence of action nodes

with several outgoing control ﬂows, loops, cross-

synchronization, or fork nodes with more than one

associated join node, but alternative implementati-

ons could be considered. An interesting issue for

future work would be to tackle the automatic mana-

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

324

Table 1: Case studies considered.

gement by means of the prototype of synchroniza-

tion structures where each fork node could have more

than one associated join node, as well as of cross-

synchronization.

5 EVALUATION OF THE

ATL-BASED PROTOTYPE

This section investigates the strengths and weaknes-

ses of our prototype. More speciﬁcally, we have app-

lied it to four different case studies and we have ana-

lysed the results in the light of several criteria regar-

ding: (1) the amount of class diagram elements that

were generated, (2) the time it took to generate the

class diagrams, and (3) maintenance aspects.

Table 1 shows the four case studies we have consi-

dered indicating the number of elements of the source

activity diagrams. The ﬁrst case study (CS1) corre-

sponds to the obesity clinical guideline used as case

study along the paper. In this case, we have used the

UML 2 Eclipse plug-in to obtain the .uml extension

ﬁle. The other three examples were determined star-

ting from the online example models proposed in (Es-

huis and Van Gorp, 2016b) (they were available in

.xmi or in .uml extension format). The case studies

we have selected correspond to: (CS2) the order and

delivery of bikes, (CS3) the handling of dermatology

patients and (CS4) the creation of a catalogue. We

have carried out several changes to these three exam-

ples to be able to apply our proposal (that is, elimi-

nating loops and forks with more than one associated

join).

We have applied our prototype to these four case

studies using a personal computer, Intel(R) Core

T M

CPU, 3.2 GHz, with 4 GB RAM, running Windows 7

Enterprise. We have performed two types of expe-

riments for each case study: (1) monitoring all the

object nodes in the model and (2) monitoring around

half of the object nodes in the model. Each type of ex-

periment has been repeated ﬁve times per case study,

measuring the response times to produce an average

time for each type of experiment and case study.

Table 2 shows the results obtained from our ex-

periments, ordering them according to the number of

monitored object nodes. It can be seen that, in each

case (all the nodes and half of the nodes), the grea-

ter the number of monitored objects the greater the

response time. Besides, all the response times are

less than half a second, with a very small difference

among them. These values, as well as the relation be-

tween the response time and the number of monitored

objects, can be seen as a good performance result.

As for maintenance aspects, our ATL-based pro-

totype would face with the evolution of business pro-

cess models better than a manual not pattern-based

approach. Speciﬁcally, every activity diagrams’ de-

sign change can be automatically processed without

software designers’ interaction.

An Eclipse distribution with the ATL transforma-

tions, together with relevant documentation regarding

the benchmark of case studies used for the evaluation

are available from (Supplementary Material, 2018).

We encourage the interested reader to try it out.

6 RELATED WORK

The need of considering data (both managed and ge-

nerated) as ﬁrst class citizens within business process

modeling has been previously tackled in the literature.

One of the ﬁrst works that exposes this need is (Ni-

gam and Caswell, 2003), which gives the notion of

business artifacts as “information chunks that the bu-

siness creates and maintains.” This concept is related

to the Occurrence notion, although this construct en-

compasses data and behavior.

Another similar approach is that of (Kumaran

et al., 2008), which deﬁnes business entity as a “domi-

nant information entity with an associated data model

and an associated behavior model in the context of a

process scope”. Although this proposal shares one of

its main objectives with our approach, it is very speci-

ﬁc since it stands up for modeling the behavior of bu-

siness entities with state machines. The Occurrence

approach is more general and does not restrict the use

of any speciﬁc technique for information representa-

tion.

There exist other papers, such as (Hull et al., 2010)

or (Künzle and Reichert, 2011), that make innova-

tive proposals for the comprehensive management of

objects and behavior. However, these works move

away from the most common standards in the dom-

ain. Our approach is demonstrated by its adaptation

to any standard (in this paper, as an example of ap-

plication, UML models are used). An example of a

work that is based on a standard is (Friedenstab et al.,

2012), which presents a BPMN extension to improve

monitoring. We note that this work also uses the term

occurrence (but in a sense very different from ours)

Generating Persistence Structures for the Integration of Data and Control Aspects in Business Process Monitoring

325

Table 2: Time results obtained for the case studies.

but it does not deal with automation aspects. In (Her-

zberg et al., 2013) an approach to use object state

transition events for reasoning about process progress

is given. In that proposal, very simple state machi-

nes are considered, while in our approach we face the

difﬁculties that come from considering parallel ﬂows

and concurrent states.

Several papers propose model transformations

for business process management. Eshuis and Van

Gorp (Eshuis and Van Gorp, 2016a; Eshuis and

Van Gorp, 2016b) automate the generation of object

life cycles (UML state machines) from business pro-

cess models (UML activity diagrams). Our propo-

sal goes a step further, since it allows the generation

of class diagrams that include intrinsically and com-

prehensively both data and ﬂow perspectives. Other

works (Br

danin and Maric, 2012; Rodríguez et al.,

2010), closer to our proposal, automatically transform

UML activity diagrams into UML class diagrams.

However, in (Br

danin and Maric, 2012) authors do

not consider alternative (decision/merge) and concur-

rent control ﬂows (fork/join), and the work presented

in (Rodríguez et al., 2010) pays special attention to

security requirements. As far as we know, our auto-

matic transformation from activity diagrams to class

models is the only one proposed within the process

monitoring context.

7 CONCLUSIONS AND FURTHER

WORK

We have presented an approach to automatically ge-

nerate storage infrastructures for monitoring tasks

that integrates control ﬂow and data perspectives.

We have shown the feasibility of the proposal using

UML Activity Diagrams, as business process mo-

deling technique, and an MDE approach, as a de-

sign strategy. After applying our approach to different

case studies, we can conclude that the results obtained

from our evaluation are promising. In particular, the

advantages of our approach could be summarized as

minimizing coding tasks, as well as simplifying the

management of the life cycle of business processes.

There are several lines of further work. At present,

as said previously, there are speciﬁc structural situa-

tions not tackled by the prototype. These situations

deserve to be investigated. Furthermore, a special re-

mark must be made regarding the pin–style modeling

of object ﬂows. Although the basic pin–style is equi-

valent to the object node style (Object Management

Group, 2015), the general pin–style notation is more

expressive allowing to handle more complex situati-

ons such as alternative pin sets. Thus, considering the

general pin–style notation is an issue of future work.

ACKNOWLEDGEMENTS

This work has been partially supported by the spanish

Ministry of Economy and Competitiveness (project

EDU2016-79838-P) and the University of Zaragoza

(project UZ2015-TEC-05).

REFERENCES

Allué, A., Domínguez, E., López, A., and Zapata, M. A.

(2013). QRP: a CMMI Appraisal Tool for Project

Quality Management. Procedia Technology, 9:664–

669.

ATL Plug-in (2017). http://www.eclipse.org/atl/. Last visi-

ted on January 2018.

danin, D. and Maric, S. (2012). An approach to automa-

ted conceptual database design based on the iml acti-

vity diagram. Comput Sci Inf Syst, 9(1):249–283.

Campanile, F., Coppolino, L., Giordano, S., and Romano,

L. (2008). A business process monitor for a mobile

phone recharging system. J. Syst. Archit., 54(9):843–

848.

Domínguez, E., Pérez, B., Rubio, A. L., Zapata, M. A.,

Allué, A., and López, A. (2017). Developing

provenance-aware query systems: an occurrence-

centric approach. Knowl Inf Syst., 50(2):661–688.

Domínguez, E., Pérez, B., Rubio, A. L., Zapata, M. A., La-

villa, J., and Allué, A. (2014). Occurrence-Oriented

Design Strategy for Developing Business Process Mo-

nitoring Systems. IEEE Trans. Knowl. Data Eng.,

26(7):1749–1762.

Eshuis, R. and Van Gorp, P. (2016a). Synthesizing data-

centric models from business process models. Com-

puting, 98(4):345–373.

Eshuis, R. and Van Gorp, P. (2016b). Synthesizing object

life cycles from business process models. Software &

Systems Modeling, 15(1):281–302.

ICEIS 2018 - 20th International Conference on Enterprise Information Systems

326

Friedenstab, J.-P., Janiesch, C., Matzner, M., and Muller, O.

(2012). Extending bpmn for business activity monito-

ring. In Proc. of HICSS’12, pages 4158–4167. IEEE.

Herzberg, N., Meyer, A., Khovalko, O., and Weske, M.

(2013). Improving business process intelligence with

object state transition events. In Proc. of ER’13, vo-

lume 8217 of LNCS, pages 146–160. Springer.

Hull, R., Damaggio, E., Fournier, F., Gupta, M., et al.

(2010). Introducing the guard-stage-milestone appro-

ach for specifying business entity lifecycles. In Proc.

of WS-FM’10, volume 6551 of LNCS, pages 1–24.

Springer.

Kang, B., Kim, D., and Kang, S.-H. (2012). Real-

time business process monitoring method for pre-

diction of abnormal termination using KNNI-based

LOF prediction. Expert Systems with Applications,

39(5):6061–6068.

Kang, D., Lee, S., Kim, K., and Lee, J. Y. (2009). An OWL-

based semantic business process monitoring frame-

work. Expert Systems with Applications, 36(4):7576–

7580.

Kumaran, S., Liu, R., and Wu, F. Y. (2008). On the duality

of information-centric and activity-centric models of

business processes. volume 5074 of LNCS, pages 32–

47. Springer.

Künzle, V. and Reichert, M. (2011). Philharmonicﬂows: to-

wards a framework for object-aware process manage-

ment. J. Softw. Maint. Evol.: Res. Pract., 23(4):205–

244.

Nigam, A. and Caswell, N. (2003). Business artifacts: An

approach to operational speciﬁcation. IBM Syst. J.,

42(3):428–445.

Object Management Group (2011). Business Pro-

cess Model and Notation v2.0 formal/2011-01-03.

http://www.omg.org/spec/BPMN/2.0/PDF. Last visi-

ted on January 2018.

Object Management Group (2015). Uniﬁed Mo-

deling Language v2.5 formal/2015-03-01.

http://www.omg.org/spec/UML/2.5/PDF. Last vi-

sited on January 2018.

Reichert, M., Bassil, S., Bobrik, R., and Bauer, T. (2010).

The proviado access control model for business pro-

cess monitoring components. EMISA, 5(3):64–88.

Rodríguez, A., de Guzmán, I. G.-R., Fernández-Medina, E.,

and Piattini, M. (2010). Semi-formal transformation

of secure business processes into analysis class and

use case models: An mda approach. Inform Software

Tech, 52(9):945–971.

Snow, V., Barry, P., Fitterman, N., Qaseem, A., and Weiss,

K. (2005). Pharmacologic and surgical management

of obesity in primary care: a clinical practice guide-

line from the American College of Physicians. Ann

Intern Med, 142(7):525–31.

Supplementary Material (2018). Generating persistence

structures for the integration of data and control

aspects in business process monitoring. Available at:

https://zenodo.org/record/807684#.WbqoAPMjG70.

January 2018.

Generating Persistence Structures for the Integration of Data and Control Aspects in Business Process Monitoring

327