Information Technology Reference
In-Depth Information
Business process management is the discipline concerned with using methods,
techniques, and software to design, enact, control, and analyze operational pro-
cesses. A large body of knowledge corresponds to process model analysis based
on model transformations. Model transformations can be reused in the context
of the abstraction problem. An example of such a transformation consists of re-
duction rule sets for Petri nets, e.g., see [4,23,27]. Each reduction rule explicitly
defines a structural fragment to be discovered in the model and a method of
this fragment transformation. Hence, reduction rule sets enable process model
abstraction through iterative rule application. As the transformed process frag-
ments are explicitly defined, each reduction rule set handles only a particular
model class. Thereby, each reduction rule set requires an argument about the
model class reducible with the given rules. The model class limits the application
of abstraction approaches based on reduction rules [5,10]. Process model decom-
position approaches are free of this limitation: they seek for process fragments
with particular properties. An example of such a decomposition is presented
in [34], where single entry single exit fragments are discovered. The result of
process model decomposition is the hierarchy of process fragments according to
the containment relation, i.e., the process structure tree. Such a tree can be
used for abstraction in process models [24]. Finally, one can distinguish model
transformations that preserve process behavior properties. In [1], van der Aalst
and Basten introduce three notions of behavioral inheritance for WF-nets and
study inheritance properties. The paper suggests model transformations, such
that the resulting model inherits the behavior of the initial model. An approach
for process model abstraction can exploit such transformations as basic opera-
tions. While the outlined model transformations can support solving the general
problem of process model abstraction, they all focus on structural and behav-
ioral aspects of models and model transformations, leaving the semantic aspect
out of scope.
Many tasks in the management of large process model collections can be traced
back to the problem of
activity matching
, which is closely related to the problem
of business process model abstraction. Examples of such management tasks are:
the search for a particular process model over a process model set or ensuring the
consistency of models capturing one and the same process from different perspec-
tives. Activity matching is realized through analysis of activity properties: activ-
ity labels, referenced data objects and neighboring activities. In [9,35] the authors
suggest activity matching algorithms and evaluate them. While the named works
explore the existing process models and do not directly address the problem of pro-
cess model abstraction, their results have a potential of being applied in business
process model abstraction. Semantic aggregation of activities relates to research
on semantic business process management. Notice that process models enriched
with semantic information facilitate many process analysis tasks, see [18]. Along
this line of research, several authors argue how to use activity ontologies to realize
activity aggregation [6,7]. It should be noticed, however, that such works imply the
existence of a semantic description for model elements and their relations, which
is a restriction that rarely holds in real world settings.
