Information Technology Reference
In-Depth Information
Perform full
analysis
Handle forecast
request
Issue report
Perform quick
analysis
abstract model, m
a
initial model, m
Perform
analysis
FA
data
Raw
data
Prepare data for
full analysis
Consolidate
results
Perform
simulation
Receive forecast
request
Collect
data
Generate
forecast report
Send
report
Prepare data for
quick analysis
Perform quick
data analysis
Raw
data
QA
data
Fig. 1.
Motivating example: initial model and its abstract counterpart
While it has been empirically shown that abstraction can significantly improve
the sense-making of large process models [25], a limited insight exists into the
criteria that experienced modelers use to decide on which activities to aggregate
into new ones. A number of techniques has been proposed that exploit
structural
properties of a process model to arrive at abstract models [5,24]. It seems likely
that experienced process modelers take a wider range of properties into account
rather than just a model's control flow. For example, the fact that two activities
use the same document and are executed by the same role may be used as
relevant inputs in deciding to cluster these two into an aggregated activity. This
situation applies, for example, to the activities
Prepare data for quick analysis
and
Perform quick data analysis
in Fig. 1.
In this paper, we complement the existing streams of work with respect to pro-
cess model abstraction by proposing an abstraction technique that incorporates
semantic aspects contained within a process model. We rely on the observation
that industrial process models are often enriched with non-control flow model ele-
ments. Examples are: data that is being processed within an activity, IT systems
invoked within particular activities, and roles assigned to activities. The central
idea in this paper is that activities associated with the
same
non-control flow
