![](bgb.png)
with Real Legacy System Logs of Table 3. That
means the business analysts agree with most of
process models structures mined from legacy, during
the incremental discovery.
5 RELATED WORK
Gunther (Gunther et al, 2008) introduced the mining
of ad-hoc process changes in adaptive Process
Management Systems (PMS). This technique
introduces extension events in the log (e.g. insert
task event, remove task event, etc) that record all
changes in a process instance. So change logs must
be interpreted as emerging sequences of activities
which are taken from a set of change operations. It is
different from conventional execution logs where
the log content describes the execution of a defined
process. The problem here is that sometimes legacy
information systems and WfMS do not generate
process change information into the log. Thus, it is
very hard to discovery process changes from these
systems. Bose (Bose et al, 2011) although,
introduced concept drift applied to mining processes.
He applied techniques for detection, location and
classification of process modifications directly in the
implementation log without the need for specialized
log containing such modifications as proposed by
Gunter. After that, Luengo (Luengo et al, 2012)
proposed a new approach using clustering
techniques. He uses the starting time of each process
instance as an additional feature to those considered
in traditional clustering approaches.
The work presented here supports the main
operations of evolutionary learning (e.g. insert and
exclusion operations) using an incremental mining
approach. Moreover, our technique does not require
extra information in the log to detect process
changes. Thus, it makes possible to avoid the
reprocessing of the complete set of logs, reducing its
total processing time.
6 SUMMARY AND OUTLOOK
This paper proposed an incremental process mining
algorithm for mining of process structures in an
evolutionary way from legacy systems. This is an
important step for the incremental legacy
modernization because it keeps the system
maintenance live while the system is modernized.
The algorithm enables the discovery of new and
obsolete relations from log as new or modified
traces are executed and recorded in the log. Thus, we
can keep all process models discovered updated with
the process definition when it changes.
In quality experiments using non-incremental
simulated data and conformance metrics, the models
discovered by IncrementalMiner present good
accuracy. Regarding the incremental approach,
IncrementalMiner also shows good precision for the
models discovered from logs with modified process
instances. During the discovery of process models
from real legacy system and also simulated logs, the
algorithm shows good results on the extracted
models (i.e. Kappa values above 0.900). Thus, our
approach could be an effective alternative for
incremental mining of process models during the re-
engineering of legacy systems.
Altogether, the main contribution of this work
was the creation of a mechanism that introduces the
incremental mining of logs with support to i) the
discovery of new dependency relations (i.e. new
tasks) and participants in order to complement a
partial or complete process model, and ii) the
identification and removal of obsolete dependency
relations in order to update an existent process
model. We also introduced an alternative way to
measure the quality of models generated during the
incremental process mining.
As future work we include the improvement of
the identification of obsolete participants in the
model (i.e. see low kappa value in simulate data of
Table 3) and the integration of algorithm and the
incremental approach in ProM tool.
REFERENCES
Kalsing, A. C., Iochpe, C., Thom, L. H.: “An Incremental
Process Mining Algorithm”. In ICEIS (1), pp. 263-
268 (2010).
Kalsing, A. C., Nascimento, G. S., Iochpe, C., Thom, L.
H.: “An Incremental Process Mining Approach to
Extract Knowledge from Legacy Systems”. In 14th
IEEE EDOC 2010, pp. 79-88 (2010b).
Cohen, J., “A coefficient of agreement for nominal scales,
Educational and Psychological Measurement”. Vol. 20
(1), pp. 37–46 (1960)
Van Dongen, J., Medeiros, A. K., Verbeek, H., Weijters,
A. J. M. M., van der AALST, W. M. P.: “The ProM
framework: A new era in process mining tool
support”. Applications and Theory of Petri Nets, pp.
1105-1116, Springer (2005)
Ma, H., Tang, Y., Wu, L.: “Incremental Mining of
Processes with Loops”. International Journal on
Artificial Intelligence Tools, Vol. 20, Number 1, pp.
221-235 (2011)
ICEIS2013-15thInternationalConferenceonEnterpriseInformationSystems
68