Information Technology Reference
In-Depth Information
this paper in discovering change points and localization of changes through a
case study. In Section 6, we project an outlook on some of the open research
questions and directions in this area. The paper ends with some conclusions in
Section 7.
2 Related Work
Over the last two decades many researchers have been working on process flex-
ibility, e.g., making workflow systems adaptive. In [4,5] collections of typical
change patterns are described. In [6,7] extensive taxonomies of the various flexi-
bility approaches and mechanisms are provided. Ploesser et al. [8] have classified
business process changes into three broad categories viz., sudden, anticipatory
and evolutionary. This classification is used in this paper, but now in the context
of event logs.
Despite that many publications on flexibility, most process mining techniques
assume a steady state process. A notable exception is the approach by Gunther
et al. [9]. This approach uses process mining to provide an aggregated overview
of all changes happened so far. However, this approach assumes that change
logs are available, i.e., modifications of the workflow model are recorded. At this
point in time very few information systems provide change logs. Therefore, this
paper focuses on concept drift in process mining assuming only an event log as
input. Concept drift refers to changes in the target variable(s)/concept induced
by contextual shifts over time [10]. While the topic is well-studied in various
branches of the data mining and machine learning community, the problem of
concept drift has not been studied in the process mining community. While
experiences from data mining and machine learning can be used to investigate
concept drift in process mining, existing techniques cannot be used due to the
complexity of process models and the nature of process change.
3 Aspects and Nature of Change in Business Processes
Three important perspectives in the context of business processes are the control-
flow, data and resource perspective. One or more of these perspectives may be
subjected to a change.
-
Control-flow/Behavioral Perspective:
This class of changes deals with the
behavioral and structural changes in a process model. Just like the design
patterns in software engineering, there exist
change patterns
capturing the
common control-flow changes [4]. Control-flow changes can be classified into
operations such as insertion, deletion, substitution and reordering of process
fragments. For example, an organization which used to collect the fee after
the processing and acceptance of an application can now change their process
to enforce the payment of fee before the processing of an application. Here
the
reordering
change pattern had been applied on the payment and applica-
tion processing process fragments. As another example, with the addition of
new product offerings, a
choice
construct is
inserted
into the product devel-
opment process of an organization. In the context of PAIS, various control-
flow change patterns have been proposed in [4,5]. Most of these control-flow
