![](bg3.png)
Figure 2: A simple carepathway process, modeled using
Petri net. After Admission, Diagnosis is performed, fol-
lowing which Treatment1 is performed in parallel with ei-
ther Treatment2 or Treatment3. Finally the patient exits the
carepathway as denoted by Release activity.
transaction type, such that for every activity instance,
there may be a schedule event, a start event, a sus-
pend event, a resume event and finally, a complete
event. Each event has a time-stamp associated with
it, which denotes the time of occurrence of a par-
ticular event. Furthermore, each event may contain
additional event specific data attributes, for example
the resource which performed the particular activity
represented by the event. Similarly, each case may
have case level data attributes, such as the gender of
the patient, age of the patient, etc. The event log is
thus a collection of a sequence of events (so-called
traces) that represent the individual cases. Event logs
extracted from the HIS can be used in a process min-
ing context for performing process discovery; or con-
formance analysis on a pre-existing or a discovered
process model.
2.2 Petri Nets
Multiple notations exist to represent process models,
for example, BPMN, UML diagrams, EPCs, Petri
nets. InterPretA uses Petri nets as a way to repre-
sent process models. The selection of Petri nets was
inspired by the virtue of the properties supported by
Petri nets which enable detailed conformance and per-
formance analysis. Petri nets support modeling of the
traditional business process concepts, such as concur-
rency, choices and sequences (Aalst, 2016). Figure 2
shows an example of a Petri net model, where places
(circles) are used for modeling logic (i.e. sequence,
choice, concurrency, loops etc.) and the rectangles
represent the activities (tasks) of the process.
2.3 Alignments
In our approach, we use alignments-based confor-
mance analysis as proposed in (Adriansyah et al.,
2011) for guiding the compliance and performance
analysis in the context of processes. As discussed
above, conformance analysis helps in determining
how well a process model fits the reality represented
by the event log. This information can be beneficial
Table 1: Example of conformance alignment moves using
Figure 2. Step 1,2,3 and 5 are synchronous moves. Step 4
and step 6 are move on model and move on log respectively.
Trace in event log a b c d e >>
Possible run of model a b c >> e f
Steps 1 2 3 4 5 6
when determining any compliance issues related to ei-
ther the complete process or some fragments of the
process. Furthermore, this information could also be
used in order to determine any performance related
problems and analysis. In this sub-section, we briefly
discuss the idea behind the alignment strategy that is
used for determining the conformance of a model and
event log as proposed in (Adriansyah et al., 2011).
Often times, the event log may contain noisy and/or
incomplete data. Alignments provide a handy way to
deal with such data. Hence, instead of relying com-
pletely on the event log, we use the information from
alignment based conformance analysis as the ground
truth. As alignments relate events in the event log
to model elements, they are ideal for the process-
oriented analysis approach supported in InterPretA.
Aligning events belonging to a trace with a process
model can result in three types of so called moves -
synchronous move, move on model and move on log.
An example alignment is shown in Table 1.
• Synchronous move: Occurrence of an event be-
longing to a trace can be mapped to occurrence of
an enabled activity in the process model.
• Move on model: Occurrence of an enabled activ-
ity in the process model cannot be mapped to the
current event in the trace sequence.
• Move on log: Occurrence of an event in the trace
cannot be mapped to any enabled activity in the
process model.
Optimal alignments provide a means to match a trace
in an event log with a corresponding model run. If
for a trace, all the moves are synchronous or invisi-
ble model moves, then that trace can be perfectly re-
played by the model.
2.4 Classification
In literature, classification techniques (Goedertier
et al., 2007; Buffett and Geng, 2010; Poggi et al.,
2013) have been applied in the field of process min-
ing, in order to address multiple problems. (Leoni
et al., 2015) provide a framework for applying classi-
fication and correlation analysis techniques in process
mining, by using the results from conformance anal-
ysis. Traditionally, classification techniques in pro-
cess mining context are used to perform tasks such as