Rule-based Business Process Abstraction Framework
Christina Tsagkani and Aphrodite Tsalgatidou
Dept. of Informatics & Telecommunications, National & Kapodistrian University of Athens (NKUA),
Panepistimioupolis, Ilisia 157 84, Greece
Keywords: Business Process Modeling, Business Process Model Abstraction, BPMN 2.0.
Abstract: Business process models have become key artifacts to represent how work is performed in organisations.
Therefore organisations maintain large process model repositories, containing complex models that are
difficult to be managed. Abstraction is a means to reduce the size and complexity of such process models and
ease the process model management. It is worthwhile noticing that several abstraction mechanisms exist in
the literature but none of them includes activities-paths, data, roles, artifacts and messages in the abstraction
procedure. To this end we present a conceptual framework that utilizes abstraction rules in order to simplify
business process models. The proposed framework is well suited for complex BPMN process models and is
built on our previous work where process abstraction rules had been introduced. This abstraction framework
is designed with focus on retaining the overall structure of the process model, but most of all it is designed to
treat different process elements (eg. data, messages, roles, etc.) other than activities as abstraction objects. A
real business case that is originated from the financial services sector is used as an attempt to exemplify and
validate the proposed mechanism.
1 INTRODUCTION
Nowadays business process management has
attracted the attention by organisations and
enterprises which focus on modelling and automating
their business processes. Business processes can be a
great source of business knowledge. Therefore, a lot
of efforts are taking place by organisations in order to
build process repositories containing hundreds or
even thousands of business process models on behalf
of different stake-holders. The way that a process
model is defined is influenced by many factors, such
as the modeler’s expertise (e.g. software engineers,
business analysts, domain experts) and the goal of
modelling (e.g. models created to provide an overall
view of process for managerial use, models defined in
a detailed way in order to be executed by technical
specialists).
After analysing diverse process models we
identified that they exhibit a number of differences:
 Differences in the way that process models
achieve the same effect or represent equivalent
unit of work
 Differences in the granularity level used to
represent the same unit of work
 Differences related to the role assigned to perform
an activity
 Differences in the control-flow relations amongst
activities between the process models
Both the heterogeneity that characterises business
process models as it is described above and the
existence of large-complex repositories, create
difficulties in process model management.
Therefore we claim that such differences can be
diminished and process management, especially
querying, can be improved, if process models are
transformed to more coarse-grained models. In this
context, abstraction mechanisms are needed that take
into consideration such differences and reduce the
size and complexity of process models by focusing on
the nature of their process elements.
Thus in this paper, we propose a conceptual
framework of a rule-based abstraction mechanism
that aims at providing a less complex view of business
processes by focusing on significant process elements
and discarding or aggregating internal process
orchestration details by applying specific rules to
activities-paths, data objects, messages, roles and
artifacts. This mechanism focuses on business
processes that are defined using the Business Process
Model and Notation (BPMN) (OMG, B.P.M., 2011)
as it is a well-known and widely used standard across
173
Tsagkani C. and Tsalgatidou A.
Rule-based Business Process Abstraction Framework.
DOI: 10.5220/0006223401730178
In Proceedings of the Sixth International Symposium on Business Modeling and Software Design (BMSD 2016), pages 173-178
ISBN: 978-989-758-190-8
Copyright
c
 2016 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved

different industries. Therefore, the rest of the paper
has been structured as follows: section 2 introduces
some concepts related to the abstraction mechanism,
section 3 describes the proposed abstraction
framework, section 4 exemplifies and validates the
proposed abstraction mechanism via a real business
case, section 5 discusses related process abstraction
mechanisms and section 6 concludes and presents our
future work.
2 BACKGROUND
2.1 Business Process Modelling
Numerous notations have been emerged into the
business process modelling space, including UML
Activity Diagrams, the Business Process Modelling
Notation (BPMN), Event-driven Process Chains
(EPCs), Workflow nets, and the Business Process
Execution Language (BPEL) that is more appropriate
for executable specifications rather than modelling
per se. From all these modelling notations, BPMN 2.0
is the prevailing standard, as it has been widely
accepted in industrial practice and we utilise it in our
abstraction mechanism.
BPMN aims at documenting and communicating
business processes between all business stakeholders.
Specifically, it is a graph-based notation — i.e. sets
of graphical symbols and rules for combining them
— for documenting flow objects, data, connecting
objects, swimlanes and artifacts. Flow objects
(Events, Activities and Gateways) define the
behaviour of business process. Data (Objects, Inputs,
Outputs and Stores) define what activities require to
be performed or produce. Connecting Objects
(Sequence, Message, Associations and Data
Associations) define the way the flow objects are
connected. Swimlanes (Pools and Lanes) represent
process participants. Artifacts (Group and Text
Annotation) are used to provide additional
information about the process. As defined by
(Chinosi and Trombetta, 2012): Process
Orchestration includes the private and public
processes of an organization. Private Processes are
processes internal to a specific organization whereas
Public processes represent the interaction between a
private process and another process or participant that
means only activities that are used to communicate
with other participants are included in the public
process. On the other hand Choreographies define the
expected behaviour that is a procedural contract
between interacting participants. Therefore a
choreography exists between pools (or participants)
as it bisects the message flows amongst them.
The current framework is focused on obtaining a
process quick view by preserving the overall process
structure. Therefore we suggest that the abstraction
mechanism leaves intact process elements that
constitute process’s choreography and abstracts only
process’s orchestration details.
2.2 Process Abstraction
Process Abstraction is a means of providing different
process views (which retain information relevant for
a particular purpose) and reducing the size and
complexity of process models by preserving essential
properties and leaving out insignificant details
(Smirnov et al., 2010). It may be applied for different
purposes such as focus on specific process model
properties (i.e. preserve pricey/frequent/long
activities), adapt process model for an external
partner, trace data/task dependencies and obtaining a
process quick view respecting ordering
constraints/roles.
Business process abstraction mechanisms should
consider different aspects (Smirnov et al., 2012): the
reason for abstraction that identifies the focus of
abstraction, the conditions that should be satisfied and
trigger the abstraction and the operations used for
abstracting a process model to a more coarse-grained
model.
Taking into consideration the above aspects the
proposed abstraction mechanism focuses on
providing process quick views based on defined rules
that describe the conditions that should be satisfied
for triggering the abstraction process and the
abstraction operation (aggregation or elimination).
3 ABSTRACTION CONCEPTUAL
DESIGN
In this section, we present the proposed conceptual
design to process abstraction. The process abstraction
mechanism is based on the use of transformation rules
that are applied to business processes in a semi-
automated way. The processes of the mechanism are
defined with the use of the BPMN standard.
The overall process abstraction conceptual design
is depicted in Figure 1 and is described in the
following.
We consider that a user creates new business
process models or modifies existing ones with the aid
of a BPMN editor and stores them to a process
repository. The process repository communicates
Sixth International Symposium on Business Modeling and Software Design
174

with the abstraction mechanism by sending existing
process models and receiving the abstracted versions
of the given process models.
Figure 1: Conceptual design of Process Abstraction
Mechanism.
The abstraction mechanism allows to apply
abstraction rules to given models, visualize process
elements candidate for abstraction, visualize statistics
of process elements abstraction, validate abstraction
and save abstraction results.
More precisely once a candidate process model of
abstraction is defined in the abstraction mechanism,
the mechanism communicates with the process
repository and the process model is loaded into it.
Then, the abstraction rules that are stored to a
repository are used to trigger process abstraction.
These abstraction rules presented in (Tsagkani and
Tsalgatidou, 2015
), are specially designed for process
models defined using BPMN 2.0 and treat activities,
paths, data objects, messages, lanes and artifacts as
abstraction objects. These rules, when triggered, use
aggregation and elimination as basic abstraction
operations. Moreover, the rules are applied to the
process model at hand while retaining the hierarchy
in Figure 2.
The process abstraction is performed in a semi-
automated way in the sense that before the
finalisation of the process abstraction based on the
applied rules, there is a validation phase where the
user is involved. All the changes that are intended to
be performed to the given process model are
visualized both graphically and statistically. More
precisely, each process element candidate for
abstraction is clearly marked; then, it is upon the user
to check its validity and accept or reject the
abstraction. The resulted abstracted process model is
also visualized along with some statistical results (e.g.
the total reduction of activities, data, messages lanes
etc.). Then, the abstracted process model is saved to
the process repository where it is associated with the
original business process model for future use.
Figure 2: Abstraction Rule Hierarchy.
4 VA L I D AT I O N
In order to evaluate the proposed mechanism a
detailed description of the ‘Loan Approval with no
Collateral’ business process coming from the
financial services industry was used. However in
order to ensure confidentiality some details of the
business process are omitted, whereas sufficient
detail is provided in order to understand and illustrate
the applicability of the presented conceptual
framework.
Firstly, process models related to the ‘Loan
Approval with no Collateral’ process were created
based on the given detailed description, using the
BPMN2Modeler that is an Eclipse-based graphical
BPMN 2.0 model Editor (BPMN2Modeler, 2013).
Therefore five process models were created (one
model for each sub-process) – ‘Requirements
Capturing and Loan Quote with no Collateral via Call
Center’, ‘Loan Request Analysis and Pre-approval’,
‘Final Loan Approval’, ‘Processing of Loan
Contractual Documents’ and ‘Loan Disbursement’.
Then the rules that are defined for the specific
abstraction mechanism were manually applied to the
process models in the order that is presented in Figure
2 and validated. Finally the abstracted models were
produced along with some statistical data that is
explained below in this section.
Due to lack of space, in this paper we use for
illustration purposes a small fragment of the ‘Loan
Request Analysis and Pre-approval’ sub-process that
is shown in Figure 3. It mainly presents the tasks
needed to perform loan pre-approval, the departments
and the roles involved, the messages that are
exchanged and the data objects and artifacts that are
associated to each task. The abstracted model that is
the outcome after applying the defined abstraction
rules of the proposed mechanism to the model
(Figure3), is presented in Figure 4.
Rule-based Business Process Abstraction Framework
175

Figure 3: Fragment of the ‘Loan Request Analysis and Pre-approval’ process model.
Figure 4: Abstracted Fragment of the ‘Loan Request Analysis and Pre-approval’ process model.
It can be observed that the two roles that co-exist
in the same Consumer Credit Department are
aggregated in the abstracted model. Also the script
task is eliminated, the user tasks that are enclosed
between the ‘Send Request’ event and the gateway
that follows are aggregated. Moreover the user tasks
that are enclosed between the ‘No Data Change’
event and the gateway that follows are aggregated.
The same applies to their data objects that are
associated with the specific user tasks. The send task
Sixth International Symposium on Business Modeling and Software Design
176

‘Inform Branch for Missing data’ and the user tasks
‘Application rejection’, ‘Application Approval’,
Inform Customer for Pre-Approval’, ‘Inform
Customer for Rejection’ and ‘Change Data’ are not
eliminated in the abstracted model as they are
associated with message exchange and are part of the
process’s choreography. Finally the artifact
associated with the user task ‘Inform Customer for
Pre-Approval’ is eliminated.
The findings from applying the proposed
abstraction framework to the ‘Loan Approval with no
Collateral’ process are presented in Figure 5 where
the total number of each process element that is
affected by the abstraction process is presented before
and after the abstraction. The findings illustrate that
the activities that constituted the process were
radically reduced from 70 to 43 that is 38.57%
reduction. Moreover data objects show a 33.33%
reduction, lanes 15.38% reduction, messages 22.22%
reduction and artifacts 100% reduction.
Figure 5: Total number of process elements before and after
abstraction.
The business case used in this section to illustrate
the value of our proposed mechanism and the findings
produced seem to fulfil our research purposes that is
to provide quick views of business processes while
retaining the overall structure of the process and
getting rid of insignificant details. Also prove that by
introducing rules to the abstraction process related to
process lanes, messages, artifacts and data objects
(besides activities) may, to a great extent, reduce the
size and complexity of the abstracted process models.
It is worthwhile mentioning that the proposed
rule-based process abstraction has been also applied
to other cases as well, with similar results. Some of
the results have been briefly presented in (Tsagkani
and Tsalgatidou, 2015
).
5 RELATED WORK
There are a number research works that analyse
business process models and base their abstraction
mechanism based on transformation rules (Bobrik et
al., 2007), (Cardoso et al., 2004), (Eshuis, and Grefen,
2008), (Günther and Van Der Aalst, 2007), (Meyer
and Weske, 2012), (Pankratius and Stucky, 2005),
(Polyvyanyy et al., 2008), (Sadiq and Orlowska,
2000) and (van Dongen et al., 2007). These research
works not only analyse different process modelling
notations and approaches but, in most of the cases,
they also concentrate on activities as abstraction
objects. An exception is the work by (Meyer and
Weske, 2012) that considers data as abstraction
objects. To the best of our knowledge the work of
(Smirnov, 2009) is the only one that analyses BPMN
1.2, but it differs from our work as it suggests rules
focused primarily on activities, while it only
describes how other elements are influenced by
activity abstraction.
The mechanism proposed in this paper targets
business processes defined in BPMN 2.0 and is
superior to other research works as it respects
process’s overall structure and proposes abstraction
rules that are focused on not only activities as
abstraction objects but data, messages, artifacts, paths
and process participants as well.
6 CONCLUSION AND FUTURE
WORK
This paper presented the conceptual framework of a
rule-based process abstraction mechanism. This
framework preserves the structure of the initial
model, focuses on process choreography and
concentrates on certain abstraction objects
(participants, activities, data objects, messages and
artifacts). Furthermore the proposed framework was
exemplified and validated using a business case
coming from the financial services industry and the
findings were discussed.
We are currently working on the detailed design
and implementation of a tool based on the proposed
conceptual framework of the abstraction mechanism.
Besides, we are planning to further validate the
proposed mechanism using more business cases.
REFERENCES
Bobrik, R., Reichert, M.U. and Bauer, T., 2007. Paramete-
13
70
99
36
5
11
43
66
28
0
Lanes Activities Data
Objects
Messages Artifacts
Before After
Rule-based Business Process Abstraction Framework
177

rizable views for process visualization.
BPMN2 Modeler Project, 2013. Eclipse.org, software
available at http://projects.eclipse.org/projects/soa.
bpmn2-modeler.
Cardoso, J., Sheth, A., Miller, J., Arnold, J. and Kochut, K.,
2004. Quality of service for workflows and web service
processes. Web Semantics: Science, Services and
Agents on the World Wide Web, 1(3), pp.281-308.
Chinosi, M. and Trombetta, A., 2012. BPMN: An
introduction to the standard. Computer Standards &
Interfaces, 34(1), pp.124-134.
Eshuis, R. and Grefen, P., 2008. Constructing customized
process views. Data & Knowledge Engineering, 64(2),
pp.419-438.
Günther, C.W. and Van Der Aalst, W.M., 2007. Fuzzy
mining–adaptive process simplification based on multi-
perspective metrics. In Business Process Management
(pp. 328-343). Springer Berlin Heidelberg.
Meyer, A. and Weske, M., 2012. Data support in process
model abstraction. In Conceptual Modeling (pp. 292-
306). Springer Berlin Heidelberg.
OMG, B.P.M., 2011. Notation (BPMN) Version 2.0 (2011).
Available on: http://www. omg. org/spec/BPMN/2.0.
Pankratius, V. and Stucky, W., 2005, January. A formal
foundation for workflow composition, workflow view
definition, and workflow normalization based on petri
nets. In Proceedings of the 2nd Asia-Pacific conference
on Conceptual modelling-Volume 43 (pp. 79-88).
Australian Computer Society, Inc.
Polyvyanyy, A., Smirnov, S. and Weske, M., 2008,
November. Reducing Complexity of Large EPCs. In
MobIS (pp. 195-207).
Sadiq, W. and Orlowska, M.E., 2000. Analyzing process
models using graph reduction techniques. Information
systems, 25(2), pp.117-134.
Smirnov, S., Reijers, H.A., Weske, M. and Nugteren, T.,
2012. Business process model abstraction: a definition,
catalog, and survey. Distributed and Parallel
Databases, 30(1), pp.63-99.
Smirnov, S., Weidlich, M. and Mendling, J., 2010. Business
process model abstraction based on behavioral profiles.
In Service-Oriented Computing (pp. 1-16). Springer
Berlin Heidelberg.
Smirnov, S., 2009, July. Structural aspects of business
process diagram abstraction. In Commerce and
Enterprise Computing, 2009. CEC'09. IEEE
Conference on (pp. 375-382). IEEE.
Tsagkani, C. and Tsalgatidou, A., 2015, October.
Abstracting BPMN models. In Proceedings of the 19th
Panhellenic Conference on Informatics (pp. 243-244).
ACM.
van Dongen, B.F., Jansen-Vullers, M.H., Verbeek, H.M.W.
and van der Aalst, W.M., 2007. Verification of the SAP
reference models using EPC reduction, state-space
analysis, and invariants. Computers in Industry, 58(6),
pp.578-601.
Sixth International Symposium on Business Modeling and Software Design
178