Page 44 - 2023-Vol19-Issue2
P. 44
40 | Andreswari, Millenia, Rizky, Haniyah & Mufti
TABLE III.
DISTRIBUTION OF CLUSTER ELEMENTS
Case ID Timestamp Activity Originator
IM0000004 07/01/2013 08:17 Reassignment TEAM0001
IM0000004 04/11/2013 13:41 Reassignment TEAM0002
IM0000004 04/11/2013 13:41 Update from customer TEAM0002
IM0000004 04/11/2013 12:09 Operator Update TEAM0003
IM0000004 04/11/2013 12:09 Assignment TEAM0003
IM0000004 04/11/2013 13:41 Assignment TEAM0002
D. Interactive Data-aware Heuristic Miner (iDHM) Fig. 5. Heuristic Miner process model
In the application of the Heuristic Miner algorithm, an
techniques and event logs to project frequency on activities
interactive Data-aware Heuristic Miner (iDHM) is used in and bindings. Activities and bindings are color coded based
ProM to generate the best dependency values automatically. on how often they occur according to the event log. The size
iDHM provides interactive exploration of the parameter space of the binding dots also scales with the frequency with which
and includes a built-in conformance check to diagnose the they occur. A small circle of the respective color to the top
quality of the found model. Thus, it is easier to explore large left of the activity to indicate a conformance issue.
parameter spaces and directly find fit problems of the model
found, for example, deviant and missing behavior in event The process model obtained shows the dependence of as-
logs. Furthermore, iDHM uses attribute data from event logs signments and status changes on open activities. There are
to reveal conditional dependencies that occur infrequently compatibility issues with update, assignment, and operator up-
[28]. date activities. This is because the Heuristic Miner trims some
data with low frequency to be displayed on the process model,
Parameters obtained from iDHM measurements include while in the process model generated by Fuzzy Miner it is
the observation frequency threshold of 0.1, dependency mea- known that the assignment has a large number of relationships
sure threshold of 0.9, binding frequency threshold of 0.1, and with other activities.
condition quality threshold of 0.5. These parameters are in
accordance with the best results which shows the accuracy E. Enhancement
of the parameters in iDHM without performing a number of
dependency variations on the test [10]. The process model is The enhancement process is carried out by analyzing the
visualized in C-net with clear semantics. originator. Based on the results of conformance, it is known
that the process model shows a tendency for workloads in
The process model produced by Heuristic Miner is in the assignment process, so an analysis is carried out on the
the form of a dependency graph (Fig. 5.). The dependency division of teams to make assignments to the process. The
relationship between two activities represents the causal de- analysis is done by knowing the division of workload for each
pendence of one activity on the other which is represented as team in the assignment process. Based on the event log data
a directed edge between the two activities. Only strong depen- used, there were 88,502 assignment activities performed.
dencies that exceed the configured dependency threshold and
are observed more frequently than the included configuration The distribution of the frequency of assignment activities
observation frequency threshold. The data perspective is taken carried out for each team shows that there is an imbalance
into account when discovering the control flow of a process. of tasks in the assignment process (Fig. 6). Team008 does
Classification techniques are used to reveal data dependen- more assignments than Team0241. This causes the services
cies between activities. These data dependencies are used to provided by Rabobank Group ICT to be less than optimal due
distinguish random noise from conditional dependencies that to the uneven workload. For this reason, it is recommended
occur infrequently. The technique is based on a control flow to pay more attention to the division of workload on each IT
perspective and ignores infrequent behavior such as noise. service desk/operations team so that the services provided
are more optimal and increase customer trust so that they can
Conditional dependencies can provide insight for process increase the amount of marketing.
analysts as they can indicate solutions and deviations from
normal process behavior. iDHM only includes conditional
dependencies whose underlying data state quality exceeds
the configured state threshold. Disconnected points represent
exclusive splits (XOR) and connected points represent parallel
splits (AND). Decision rules can be filtered based on the
decision rule quality threshold. C-Net conformance checking