Process discovery algorithms automatically construct a process model based
on the behavior that was observed in the event log, and in the ProM framework
there are many different discovery plug-ins available. However, the nature of our
test log poses some challenges. On the one hand, there are only a few process
instances available, which at the same time are very long, and contain logged
tests that are executed in different phases of the test process (cf. ‘multiple’ codes
in Section 4.1). On the other hand, we already know that the process is very
flexible (as parts of it might need to be redone depending on the outcome of the
performed tests), while most of the traditional discovery algorithms described in
literature assume that the underlying process is “structured” (i.e., the number
of possible paths through a process is limited or the paths have some regular
form). Most process discovery plug-ins in ProM also focus on structured processes.
Fortunately, there are also several plug-ins that are able to deal with less
structured processes [2, 31].