International Conference on ICT and Knowledge Engineering
Using Inductive Miner to Find the Most Optimized Path
of Workflow Process
School Information Technology
Graduate School Information Technology
School Information Technology
School Information Technology
Abstract—Nowadays dealing with the concept
of customer’s loyalty and customer’s satisfaction has allocated significant importance to itself. All businesses try to make their customers as satisfied as
possible in order to create and increase the extent of their brand awareness
and brand image/value. Accordingly, they consistently keep integrating their businesses
with technology to help them improve their business through data mining or
performance checking techniques and approaches. In this study, the main focus of the target organization is to
create a web-based application with the
intention of managing job-related tasks and linked data to
another web-based application in order to reply
customer inquiries about the progress status of the offered service in real-time.
Even if a sample
Company A has supported applications, the company still faces problems and
challenges from those departments in which they are dealing with job-related tasks especially customer service department. The customer service department cannot always follow and track the
exact job-related status of the tasks on a
web-based application in a real-time manner.
Typically, there would
be too much waste of time for an employee to find a job status manually (i.e., such as making a call and searching for those employees whose
responsibility on that job is accomplished one by one) when receiving a call from customers. To find out what happens within the workflow processes, the
researchers in this study have intended to use process mining techniques so as
to analyze an event log previously collected from a web-based application which supposed to manage and control job-oriented tasks.
After contemplating on
various methods and algorithms supported by process mining approaches, the
authors of the study decided to choose Inductive Mining technique (i.e., Petri net-based graphs and models) to find out the workflow of the ongoing tasks as well as
investigating the performance of each flow in detail.
978-1-5386-2117-2/17/$31.00 ©2017 IEEE
Inductive mining, Patri net
most of organizations try to create their own information technology to support
and facilitate the processes in their business processes and organizational
The in-house information technology has been created for supporting
business processes, which usually make the data more accurate while reducing
the total response rate to deal with customers’ demands, questions, concerns, difficulties and Q, etc. Even though organizations try to use the state-of-the-art technology to support their business progress the
organizations yet face problems getting along with call center tasks, customer
complaints and so on challenges.
technology is not necessarily able to reduce the total response time needed in
order to address customer satisfaction unless the business is endangered by
less satisfied and more aggressive type of customers who might be troublesome
for the reputation and image of any organizations.Process mining 1 techniques can be helpful to analyze work flow processes of
transactions within an event log generated through an Information system. Different algorithms can lead to generation of different results
that depending on the type and specifications of processes. The results and outcomes of the process mining techniques can be
used to show the existing errors or redundant processes that fixing them can
significantly reduce the efficiency of the workflow process in terms of time
the data that is used as an input to a process mining platform need to conform
with at least one of these 3 contributions: (1)
It should include “Activity Type”, since it is crucial to know what
activity has been excecuse and performed in each process, and as a result set
and define it as “Case ID”. (2)
Date and time of the
occurrence of the relevant activity, since it is important to know when the
activities have happened in each process, in what order or sequence they have
been executed, and as a result set and define them as “Time Stamp”.
(3) The person or system who has been responsible or in charge of
performing or creating each relevant activity, and eventually define and set it
research presents a novel approach to apply process mining through Inductive
Miner 2 (Petri net-based) algorithm to simulate workflow process of what happened in a
medical system by comparing each trace of the event log with the master process
model in order to find any violence of the rules or any bottlenecks and
problems within the existing system. In
this research an event log file from a private organization was used and the
results showed that process mining can precisely help the administrators of the
organization to analyze real-time workflow processes of what
happened using a web-based application and finally
addressing the problems or discrepancies with the purpose of reducing the total
response time needed to offer a service to customers. The proposed approach in this study resulted in increasing the
extent of customers’
the section in that organization, as well as better management of the personnel
needed or ignored in different parts of the outpatient service center.
mining is a technique which is capable of creating models from event logs. The models can be discovered based on the real processes captured
by an information system.
Depending on the type
of the event logs and the type of the business environment or situation,
process mining supports a wide variety of algorithms in which they can be
applied on varied types of data and sources. The most interesting aspect of the process mining is that the
event logs can be used in order to create and generate process models/graphs that are compatible with the real-life situations and scenarios to some degree depends on the type
and accuracy of the algorithm applied within the process discovery process.
the most popular and commonly used process mining platforms is Disco, which is
an open source software developed by Fluxicon company. Disco has a user-friendly and straightforward
environment, making it understandable for most of the users who have ever tried
it. Disco uses Fuzzy Miner algorithm in
order to analyze the imported event-logs
leading to generation and simulation of the fuzzy process models in which can
be investigated in terms of:
(1) frequency or number of times and actions has been performed/executed, (2) duration of the time allocated between the performed tasks and
However, most of the
time, researchers use Disco in order to pre-process, cleanse, or filter the data based on their preferences
feature of Disco is ability to convert the type of the imported data into other
formats such as XES and MXML supported by ProM, which is another popular
process mining platform.
Inductive Miner2(Patri net)
Miner is an improvement on the Alpha Miner and Heuristic Miner algorithms that
aim to produce a sound process model. The
Inductive Miner approach is based on
finding a prominent split in an event log firstly, and then by detecting how
these splits are related to each other, secondly, and then by connecting the
splits logs together, thirdly.
phase, researchers usually get the event logs previously captured and collected
from a versatile variety of web-based applications in different
formats or structures such as .CSV. In this study, after importing the collected CSV file into the
Disco environment, a pre-processing (so-called cleansing) step was done in order to select the appropriate type of the data
needed throughout the process discovery stage, while getting rid of the
unwanted and redundant types of the existing data within the event logs.
1. Importation of CSV file
into Disco environment
after the data importation and cleansing steps, the Case ID, Activity Type,
Timestamp and Resource columns were identified and selected.
2. Timestamp filtering
3. Endpoints filtering
2 types of filtering
(i.e., based on Timestamp and Endpoints) were run in Disco in order
to investigate and keep in track only the desirable and useful parts of
data prior to exporting theminto the
Rapid Miner studio environment.
4. Importing XES file into
Rapid Miner environment
5. A screenshot of the
generated model obtained from Inductive Miner algorithm in RapidMiner platform
6. Model from Inductive
Miner algorithm(Peri net)
results obtained from the Inductive Miner algorithm showed the existence of
four different paths of workflow.
By comparing the
generated paths with the performance graphs generated in Disco, more insightful
information was discovered about the workflow processes in detail.
7. Workflow simulation of
the processes generated by Disco fuzzy miner algorithm (left: time performance based,
right: frequency performance based) as shown in the following box
contemplating on customer’s service in token “J04”, it is obvious that the assigned
tasks have been completed in a very slow and time consuming manner
Comparasion of result
infrastructure of the web-based application was designed in
such a way that be capable of showing the links of data between two running web-based applications in real-time. The web-based application also was capable
of controlling the ongoing works and tasks using mobile technology in order to
keep the status of the jobs/tasks updated in real-time (i.e., the mobile technology identified
and marked the status of the ongoing jobs and tasks with “J03”
throughout the process
In other words, the
developed web-based system allowed the
administrators to skip the need to follow up the latest status of tasks/jobs manually in case of the mobiles are run out or in case any
loss of the signal was happening.
The designed web-based application allowed the administrators to follow up the
recent updates in regard to the status of “J03”
adding job details directly without syncing the job from the mobile devices. Accordingly, one of the advantages of the developed system is that
each user can keep manually using their mobile device, why the administrators
from the headquarter operating office could track and monitor each movement and
behavior of the users in an online and real-time based behavior.
Overall, the data
collected data from the developed web-based
application included 5,508 cases in a month while only 1,611 cases showed
tendencies to follow up the status job “J03”. Accordingly, many of the cases have skipped the “J03”
The results of the study could manifest and
discover the performance of the activities in each workflow in a more detailed
and straightforward approach.
Based on the findings
of the research, the those kind of the routes of the workflow that passed
through “J03” always had a shorter period than the ones in which were skipped. Based on the organization’s
policies and service standards all of the tasked need to include the process “J04”
and need to be
completed in 24 hours.
If a task follows a
path that includes mobile application, then the following jobs/tasks status would be updated in real-time by the help of the developed web-based application in the organization’s call center.
By applying this
approach, the problems of the customer service department will be decreased and
the duration of the response time to each customer will be much shorter and
improvements will affect and improve the organization’s brand’s image and reputation in a more
to the results of the study, it was understood and discovered that overall the
system workflow is well related to each job-related employee, therefore the degree of support was quite
The only problem and
challenge ahead was that due to the management’s resistance to change, it is not always an easy task to convince
the managers and administrators of the organizations and companies to use and
apply a novel system similar to the one was developed and proposed in this
study. Unfortunately, managers are willing
to be stuck to using traditional systems and approaches, and therefore, in
order to motivate them to use a new system like the one discussed in this
paper, there is a need to show them the real advantages and benefits of the
novel approaches to track and monitor the behaviors of the personnel in real-time situations and scenarios.
the results of the research showed that the performance of the existing
workflows will surely get better if every worker, who is in charge of an
assigned job, followed every step to complete that job in a more timely manner. Flexibility of the developed system should be used only for urgent
cases (J03 Should be skip in urgent case
such as when got problems with mobile server, the weak of mobile signals, or
ect.).If all of the users carefully and
attentively follow every step in the designed system, they can decrease the
duration of the long waiting time by at least 56% leading to a better efficiency and higher customer retention and
satisfaction rates, as well as improvement of the brand image. Following all of the mentioned steps in job-tasked controlling in a web-based
environment would make real-time status to call center web-based and decreased competing with those departments. If they work following the right path even it has 1 step more than
another it can decrease time to filled data in system. Job status will appear in another web-apps in real-time then workers need not to
followed the job/task status by manual.
less conflict between 2 department.
As a result, there is
a need to enforce a more intense policy in order to force the users to fully
use the system as much as possible for improving the performance of the company
by not wasting any investment budget on development of lousy or impotent
der Aalst. Process Mining: Discovery, Conformance and Enhancement
of Business Processes. Springer-Verlag, Berlin, 2011.
Inductive Miner online https://www.futurelearn.com/courses/process-mining/0/steps/15642Cited: October 7, 2017.
3 W. Premchaiswadi, 2015,Process
Mining, Engineering Journal of Siam University, Vol.16,
Issue 1, No.30, pp.1-3.
4 Porouhan, P., & Premchaiswadi, W. (2017).
Pattern mining and process modelling of collaborative interaction data
in an online multi-tabletop learning environment. International Journal of
Knowledge Engineering and Data Mining (IJKEDM) , 4 (2), 1-31.
5 Porouhan, P., & Premchaiswadi, W. (2017).
Process Mining and Learners’ Behavior Analytics in a Collaborative and
Web-Based Multi-Tabletop Environment. International Journal of Online Pedagogy
and Course Design (IJOPCD) , 7 (3), 1-25.
& Porouhan, P. (2015). Process modeling and bottleneck
mining in online peer-review systems. SpringerPlus (Computer Science) , 4 (441), 1-18.
& Porouhan, P. (2015). Process modeling and decision
mining in a collaborative distance learning environment. Decision Analytics , 2 (6), 1-34.