Page 41 - 2023-Vol19-Issue2

37 |
Fig. 1. Modeling flowchart
A. Heuristic Miner
Heuristic Miner is an algorithm in process mining that fo-
cuses on control flow to generate process models in Heuristics
Nets format for certain event logs. The starting point of the
heuristics miner is the construction of a dependency graph
[12]. Heuristic Miner can handle noise and can be used to
express large events in the event log. Heuristic Miner only
considers the sequence of events in a case without considering
the sequence of events between cases. Event data is sorted
by timestamps from earliest to newest. Heuristic Miner ex-
tracts a C-net process model, with a finite set of activities and played like small roads on the map. Several small cities that
dependency relationships from those activities [22].
The steps used in forming the process model with the
Heuristic Miner algorithm include:
1. Create a dependency graph matrix to store the number
of dependency relationships between two activities.
2. Determine the Dependency threshold, Positive Obser-
vation threshold, Relative to best threshold as the basis for
selecting the dependency relation that will be raised in the
process model.
3. Checks if there is a short loop (length-one-loop or
length-two-loop).
4. Determine parallel relations (XOR or AND) between
activities.
5. Process models can be formed [22].
The Heuristic Miner algorithm can handle spaghetti pro-
cesses problems by focusing on the number of occurrences of
Andreswari, Millenia, Rizky, Haniyah & Mufti
relations between activities with one another in the event log
in forming the process model. Spaghetti processes are event
logs that have too many complicated relationships making
them difficult to read. The relationship between activities
with a small number will not appear in the process model
formed by the Heuristic Miner algorithm. The advantage of
the Heuristic Miner algorithm is that it can display repetitive
activities in the form of length-one-loop and length-two-loop
if the event log contains these relations. The results of the pro-
cess model can be converted into a Petri net that can display
parallel relations such as XOR and AND. The disadvantage of
the Heuristic Miner algorithm is that the number of relations
between activities that have a small occurrence in the event
log can be considered a disturbance so that these relations
cannot appear in the process model [22].
The implementation of the heuristic miner is to analyze the
software structure and implementation information. Source
code repository mining enables an understanding of how de-
velopment teams work together during software system de-
velopment and makes it possible to perform historical and
evolutionary analysis of source code elements and structures
[23].
B. Fuzzy Miner
The Fuzzy Miner Algorithm is an algorithm that is intended
for complex real-life log data. The process model generated
at the discovery stage can be simplified by the Fuzzy Miner
algorithm to avoid spaghetti processes through the simplifi-
cation process by paying attention to the right variables to
get good conformance [10]. In its implementation, the Fuzzy
Miner algorithm uses the concept of a road map as a parable
to visualize the resulting model. In the concept of a roadmap,
components that have a less important role will not be dis-
are still in one part of a big city will be combined, so the right
decision criteria are needed as a basis for the simplification
and visualization of a process model [24].
There are two basic matrices used, namely significance
and correlation. Significance can be determined for activities
(nodes) and relationships (edges) by measuring the value of
relative importance so that the level of importance can be
determined. Correlation is used to determine the relationship
that takes precedence over activities by measuring how close
the relationship between two activities is [25]. In detail, the
measurements of significance and correlation include:
1) Log-Based Process Metrics: In measuring significance,
there are two categories: unary significance and binary sig-
nificance. The significance that is measured for activities in
the process is called unary significance, while the significance
that measures the relationship in the process is called binary
significance. Correlation is often called ”binary correlation”