What is the definition of data mining that the author mentions?? How is this different from our current understanding of data mining? What is the p

  

 

What is the definition of data mining that the author mentions?  How is this different from our current understanding of data mining?
What is the premise of the use case and findings?
What type of tools are used in the data mining aspect of the use case and how are they used?
Were the tools used appropriate for the use case?  Why or why not?

Don't use plagiarized sources. Get Your Custom Essay on
What is the definition of data mining that the author mentions?? How is this different from our current understanding of data mining? What is the p
Just from $10/Page
Order Essay

DataminingWeek5CaseStudy.pdf

Week5Questions.docx

Technology in Education – Research Article
Educational data mining using cluster analysis and decision tree technique: A case study
Snjez?ana Kriz?anic´ 1
Abstract Data mining refers to the application of data analysis techniques with the aim of extracting hidden knowledge from data by performing the tasks of pattern recognition and predictive modeling. This article describes the application of data mining techniques on educational data of a higher education institution in Croatia. Data used for the analysis are event logs downloaded from an e-learning environment of a real e-course. Data mining techniques applied for the research are cluster analysis and decision tree. The cluster analysis was performed by organizing collections of patterns into groups based on student behavior similarity in using course materials. Decision tree was the method of interest for generating a representation of decision-making that allowed defining classes of objects for the purpose of deeper analysis about how students learned.
Keywords Educational data mining, cluster analysis, decision trees, case study, log file
Date received: 30 September 2019; accepted: 18 January 2020
Introduction
Data mining is a widely spread approach for analyzing
large data repositories to extract necessary or useful infor-
mation. The goal of data mining application is to extract
hidden data patterns and to detect relationships between
parameters in a vast amount of data. The exploration of
data in education using data mining techniques is com-
monly known as educational data mining. 1
Different edu-
cational data are stored in large databases. This is
especially true for online programs, for the support of
teaching processes and in which student learning behaviors
can be recorded and stored. The most common type of such
information systems is learning management system. 2
Many educational institutions evaluate the performance
of their students based on final grades which depend on a
course structure assessment and learning objectives to
achieve an effective and consistent learning process. 3
In this article, cluster analysis and decision tree tech-
nique are used to analyze student behavior for a real
e-course during one semester. The data used for analysis
are event logs downloaded from an e-learning system for
one e-course at a higher education institution in Croatia for
a student generation in 2017/2018. The file in which infor-
mation system records are stored is called a log file and the
data in it are called event logs. 4
Cluster analysis is a technique for creating organized
collections of patterns into groups based on their similarity
of some property or action. 5
Because of the fact that cluster
analysis is used for different purposes in educational data
mining, one of the most interesting areas of its application
is for grouping the students to identify typical patterns of
behavior. 6
1 Faculty of Organization and Informatics, University of Zagreb, Varaz?din,
Croatia
Corresponding author:
Snjez?ana Kriz?anic´, Faculty of Organization and Informatics, University of
Zagreb, Varaz?din 42000, Croatia.
Email: [email protected]
International Journal of Engineering Business Management
Volume 12: 1–9 ª The Author(s) 2020
DOI: 10.1177/1847979020908675 journals.sagepub.com/home/enb
Creative Commons CC BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(https://creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without further
permission provided the original work is attributed as specified on the SAGE and Open Access pages (https://us.sagepub.com/en-us/nam/
open-access-at-sage).

https://orcid.org/0000-0002-0834-1710
https://orcid.org/0000-0002-0834-1710
mailto:[email protected]
https://doi.org/10.1177/1847979020908675
http://journals.sagepub.com/home/enb
https://creativecommons.org/licenses/by/4.0/
https://us.sagepub.com/en-us/nam/open-access-at-sage
https://us.sagepub.com/en-us/nam/open-access-at-sage
http://crossmark.crossref.org/dialog/?doi=10.1177%2F1847979020908675&domain=pdf&date_stamp=2020-02-25

The purpose of decision trees is to identify specific
object classes. Decision trees use different object attributes
to classify different object subsets and do not use just one
attribute or a fixed set of attributes. 7
The attractiveness of
decision trees is in their easiness for understandability and
interpretability.
The aim of this article is to investigate which recorded
elements of student behavior in the e-learning system could
contribute to successful passing of exams in the observed e-
course. The research questions this article is trying to
answer are as follows:
1. Which student information can be extracted from
event logs of an e-learning system?
2. Which variable values have a significant influence
on grouping students with regard to their behavior in
the e-learning system?
The motivation for writing this article comes from
finding a course that is interesting to analyze due to its
variety of student activities based on which advanced data
mining techniques can be applied to improve content
management in that course. The quality of e-course exe-
cution at higher education institutions in Croatia reflects
the quality of teaching according to which higher educa-
tion institutions are ranked.
In the literature review, an analysis of the existing lit-
erature is conducted. In this chapter, educational data min-
ing, application of logs, cluster analysis, and decision tree
technique are researched. Further on, research methodol-
ogy of this article is presented with the aim of introduction
on research data and research technique. Methodology is
followed by a description of the results obtained by cluster
analysis and decision tree technique. Article ends with
final discussion remarks on perceived knowledge and
future work.
Literature review
Logs could contain a wide range of information about pro-
cess executions. 8
Data mining shares some characteristics
with automatic process discovery techniques, and in data
mining, “meaningful information is extracted from fine-
granular data, so that these techniques of automatic process
discovery are subsumed to the research area of process
mining.” 4
Data mining is the process of extracting useful informa-
tion and knowledge from a large set of data warehouses. It
involves the application of data analytics tools to detect
unknown patterns and relationships in large data sets. 1
“Data mining is a multidisciplinary area in which several
computing paradigms converge: decision tree construction,
rule induction, artificial neural networks, instance-based
learning, Bayesian learning, logic programming, statistical
algorithms, etc.” 9
In addition, some of the most useful data
mining tasks and methods are statistics, visualization,
clustering, classification, and association rule mining.
These methods reveal new, interesting, and useful knowl-
edge based on the available information. 9
The application of data mining techniques on educa-
tional data is called educational data mining. 6
The primary
goal of using data mining techniques in the field of educa-
tion is to develop models by which we can predict the
overall performance of students in selected courses. 1
The steps to improve the level of education are as
follows:
? Creating data sources of predictive variables. ? Identification of different characteristics or factors
that influence student learning performance during
academic life.
? Construction of a predictive model using classifica- tion data mining techniques based on predictive
variables.
? Validation of a model that was developed according to students’ performance while learning.
10
As there are many databases containing students’ infor-
mation, it is possible to operate with large repositories of
data reflecting how students learn. 11
Folino et al. were
investigating the usage of external-memory decision tree
induction approach to deal efficiently with large logs. 8
Data
mining techniques economically provide adjustable educa-
tion, effectively improve the system, and reduce the costs
of an educational process. 10
Higher education institutions
are concerned about the quality of education and use a
variety of ways to analyze and advance understanding of
student achievements. 3
In the context of teaching and learn-
ing, student data can be used to create and construct pre-
dictive models through which student performance can be
identified. 3
“By extracting information from data, it is pos-
sible to generate process models representing various pro-
cess scenarios in education.” 11
Asif et al. state that the aim
of forecasting in educational data mining is to predict stu-
dents’ educational outcomes. 6
Examples of data mining
techniques usage in the e-learning process are assessing
student learning performance, ensuring course adjustment,
and generating learning recommendations based on student
behavior while learning, evaluating teaching materials and
educational courses, providing feedback to teachers and
students, and discovering atypical student behavior while
learning. 9
Ma´rquez-Vera et al. present a method for predicting
student success, which consists of the following commonly
used steps in educational data mining:
1. Data collection. Refers to collecting all available
student information. Users create data files starting
with e-learning databases. 9
2. Data preprocessing. At this stage, a data set is pre-
pared for the application of data mining techniques.
To successfully complete this stage, data
2 International Journal of Engineering Business Management

preprocessing methods such as data cleansing,
variable transformation, and data partitioning must
be used.
3. Data mining. Data mining algorithms, such as clas-
sification and clustering, are applied to predict
student success.
4. Interpretation. At this stage, the models are ana-
lyzed to predict student success. 12
Various data mining techniques such as classification
and clustering are applied to reveal hidden knowledge
from educational data. 6
Clustering is used by pattern anal-
ysis, decision-making, and machine learning, which
includes data mining, document retrieval, image segmen-
tation, and pattern classification. 5
Various pieces of infor-
mation stored for each event can be used for clustering,
correlating, and finding causal relationships in the event
logs. 4
Using cluster analysis, we separate students into
groups, so that students in the same group share the same
progression within the group. 6
Data clustering used with
k-means algorithm enables teachers to predict student
performance and associate learning styles of different
learner types and their behavior with the aim of collec-
tively improving institutional performance. 13
K-means
is the most popular and the simplest partitional algori-
thm used for clustering. 14
“Measuring the similarity of
two objects is done by calculating a distance measure
such as the Euclidean Distance attributes having numer-
ical values.” 6
Several methods have been developed to solve classifi-
cation problems. Among all these methods, decision tree is
recognized as suitable, because it is considered to be one of
the most commonly used methods in the supervised learn-
ing approach. 15
Decision tree is a classification algorithm that is dis-
played in the form of a tree in which two different types of
nodes are connected by branches. 3
The induction of the
decision tree is done through a supervised knowledge dis-
covery process in which prior class knowledge was used to
channel new knowledge. 16
The tree consists of internal
nodes that match the logical attribute test and the connect-
ing branches which represent the test outcomes. 6
The deci-
sion tree classifies instances by sorting them down the tree
from the root to the leaf nodes. 2
The decision tree is con-
sidered to be a procedure that decides whether a particular
value will be accepted or rejected, uses IF-THEN rule, and
ensures that the current state is mapped to a future state to
make a different decision. 3
IF-THEN rule is one of the
most popular forms of knowledge representation because
it is easy to understand and interpret by nonexpert users
and can be directly applied in the decision-making pro-
cess. 12
The nodes and the branches form a consecutive
path through the decision tree that reaches the leaves, and
it represents a specific mark. All the nodes in the tree
correspond to a subset of data. Ideally, the leaf is clean,
which means that all elements in the leaf have an equal
chance of being a target variable or a class. 6
In the context
of learning through the decision tree, the target variables
refer to attributes. Each attribute node splits a set of
instances into two or more subsets. The root of the tree
corresponds to all instances. 17
Decision trees are easy to understand and well adapted
to the classification problems. They suffer from a sensi-
tivity of the data used in their construction and they are a
less natural model for regression. The advantage of deci-
sion trees is that there is a large number of efficient algo-
rithms, which can find approximately optimal tree
architectures. 18
In addition, decision trees are able to
break down the complex problem of decision-making into
several simpler ones. 15
The steps in decision tree building are as follows:
1. Suppose C is a set of objects to be classified by
starting from the current node. If all members within
a set C are of the same class or C is empty, we
determine that the current node is a node of the leaf,
label it according to its class, and complete the pro-
cedure. Otherwise, we move on to step 2.
2. Suppose Ai is the attribute selected for the current
node. The attribute Ai has possible values in Vi ¼ fAi1, Ai2, . . . , Aivg.
3. We use attribute values to divide the set of objects C
into mutually exclusive and exhaustive subsets fCi1, Ci2, . . . , Civg. Each subset of Cij contains objects in C which have the value Aij for the attribute Ai.
4. We create a child node in the tree for each attribute
of the Aij value and the corresponding subset of Cij.
Then we label the arc from the current node to the
child node with the attribute value Aij.
5. For each child node, we recursively call the pro-
cedure over the subset Cij with the set of available
attributes fA ? Aig.7
Decision nodes are usually represented as squares and
child nodes are drawn to the right of their parents. 19
The
decision tree can be used to predict and classify new stu-
dents depending on their activities and decisions made,
because the attributes and values, which are used for clas-
sification, are also represented in the form of a tree. 9
According to knowledge from the data associated with the
execution of numerous traces, the aim is to build a decision
tree model for use to predict membership into the clusters
for forthcoming enactments. 8
In comparison with other
data-driven approaches, decision trees are easy to under-
stand and their application does not include complex com-
puter knowledge. 20
Methodology
In this paragraph, research methodology used for conduct-
ing the analysis will be presented. First, the proposed model
for educational data mining using cluster analysis and
Kriz?anic´ 3

decision tree technique is presented. Then, the data source
and the data type are described.
Educational data mining model
According to the literature researched in the previous stage,
the activities shown in Figure 1 are recognized as some of
the most important ones in educational data mining using
cluster analysis and decision tree technique.
First, the analyst needs to select a data set to analyze,
that is, to select the targeted e-course. After selecting an e-
course, log files from an e-learning environment need to be
downloaded. On the basis of the downloaded event logs,
the next phase of the educational data mining process can
be provided. When the data are downloaded and stored,
data cleaning activity can be launched. In this activity, the
data analyst performs unnecessary data cleaning and data
separation of information that are not relevant for the anal-
ysis. After data cleaning activity, data partitioning is per-
formed. This means that the relevant data are extracted and
combined for further analysis. This activity depends on
data mining techniques and the outcome of the analysis.
Once there are manageable data, the application of cluster
analysis can be performed to create groups of students
similar within the group and different to another group.
According to these groups, it is possible to apply another
data mining technique over the obtained data, for example,
decision tree technique. In other words, after having the
obtained data from cluster analysis, the same could be
exported and prepared for decision tree technique
implementation. When there is a model resulted from the
previous activities, the model validation can be performed.
The analyst should be informed in a way of controlling the
correctness of the resulting model. After confirming the
model validation, the obtained model can be interpreted
according to the results.
Data description
The data used for the analysis are event logs downloaded
from an e-learning system for one e-course of a higher
education institution in Croatia for a student generation in
the 2017/2018 academic year. The time span in which the
data were observed was from February 2018 to June 2018.
Originally, there were 62,985 records, and after data clean-
ing and removing around 3000 records about course admin-
istrations and teachers, 59,605 records remained for
analysis. These records represented the raw data which
consisted of access date and time, student names, context
(e.g. lecture materials), component (e.g. “record”), activity
description, source (e.g. “web”), and the IP address of the
student who accessed the e-course.
The data cleaning included removing information about
the activity of system administrators and teachers because
only students’ behavior in the e-learning system was inter-
esting for this analysis. In addition, due to the sensibility of
the data and privacy, only a subset of anonymized data was
extracted for further analysis. In total, there were 185 stu-
dents participating in the e-course during the semester.
There were two mid-term exams which were performed
Figure 1. Educational data mining process using cluster analysis and decision tree technique.
4 International Journal of Engineering Business Management

in April 2018 and in June the same year. Each mid-term
exam had 40 points at maximum, and there was no thresh-
old for the required minimum points. The results of the
mid-term exams were assigned for each student individu-
ally in the e-learning system.
As stated in previous research, 11
the following variables
were recognized as significant for cluster formation:
1. “Context” from the event logs that provides infor-
mation about the e-content type.
2. A description of the activity that relates the activity
with the unique student identification label.
Previous research aimed to find groups of students
according to their behavior in the e-learning system but
another generation. By applying the same variables on
another data set (the generation 2017/2018 in this case), the
usefulness of the context variables is tested. To further ana-
lyze and understand student behavior, this study takes a
deeper approach and applies additional decision tree tech-
nique on data.
The values of the variable “Context” were as follows:
access to lecture materials, access to auditory materials,
access to laboratory materials, and access to forums. Lec-
ture materials were available to students each week when
the teaching topic was processed. Before or after the
lectures, students were able to download the teaching mate-
rials from the e-learning system. Before auditory exercises
(AEs), students were able to download and print teaching
materials so they could easily follow the class. On average,
it took about five clicks to download each material. Labora-
tory exercises (LEs) were held in laboratory classes at a
higher education institution where students were asked to
show independency in solving the assignments. During the
class, students were required to download e-learning mate-
rials, which also required approximately five clicks. The
forums consisted of a Discussion Forum, where students
were able to ask questions about the e-course and commu-
nicate mutually, and a News Forum that contained news
related to the e-course and teacher consultations, which
were addressed by the teachers themselves.
After data cleaning, a pivot table was created, contain-
ing information about frequency of access for each student
according to his or her recorded identification label. Fre-
quency of access to the e-content shows the popularity of
the content, and the “popularity” can be measured by how
many times requests are made for the e-content during the
semester. 21
By the frequency of access to the e-content in
the e-course, it is possible to determine which e-content
students recognized as relevant for passing the mid-term
exams and whether the frequencies of the access influenced
the final outcome of the exams. 11
So, the pivot table con-
tained student identification labels in a form of numbers
and numerical frequencies of access to materials from lec-
tures, AEs, LEs, and forums for each student. This table
was imported into RapidMiner 22
tool that has been used for
performing the next data mining techniques: cluster analy-
sis and decision tree. These data mining techniques were
selected because, according to the literature, 12
data mining
uses a more direct approach, such as the percentage usage
of well-classified data, while statistical techniques are usu-
ally used as a quality criterion for the veracity of the data
given model. Besides, data mining techniques work well
with very large amounts of data, while the statistics does
not work well in large databases with high dimensionality.
The tool settings for the cluster analysis were the applied
algorithm was k-means, the number of groups was 3
(according to testing, it was considered to be the best value
with promising results), the grouping variable was stu-
dent’s ID, the method chosen for normalization was
Z-transformation, measure types for grouping were
numerical measures, and chosen numerical measure was
Euclidean distance. Finally, the selected influential vari-
ables on grouping were frequencies of access to materials
from lectures, AEs and LEs, and forums.
The tool settings for the performance of decision tree
technique were respectively: the target variable whose out-
come was intended to be predicted is the number of stu-
dents’ points achieved in two mid-term exams where both
mid-term exams amounted to 80 points in total. Student’s
points are the variable that yields the highest information
gain. Further on, the method chosen for normalization was
Z-transformation, the criterion by which the decision trees
were created was the least square, maximal depth of the
trees was 10, minimal leaf size was 2, minimal size for split
was 4, and a number of prepruning alternatives was 3.
These settings were applied to all decision trees which
resulted from this research. The difference was in the size
of the minimal gain, and it was as follows:
? For the decision tree of the cluster number 0: 0.105. ? For the decision tree of the cluster number 1: 0.081. ? For the decision tree of the cluster number 2: 0.08.
These values were chosen considering the best resulted
branching of the trees and the acceptability of the results for
interpretation according to previously obtained clustering
models.
Results
The educational data mining analysis, conducted in this
research, resulted with one model by cluster analysis show-
ing groups of students according to their behavior in the e-
learning system and three models of decision tree made
according to previously conducted cluster analysis. The
following section describes the results of the grouping anal-
ysis and decision tree. In addition, a box plot diagram made
by points of the students from the mid-term exams is pre-
sented to show the verification of gained models by stu-
dent’s success.
Kriz?anic´ 5

Interpretation of the grouping results
The aim of grouping the students was to find groups of
students who were similar to each other within the group
and different in respect to the other groups. The similarity
depends on the behavior of the students in an e-learning
system during the semester. Behavioral intention is an
important predictor of student behavior that varies between
different behavioral, control, and normative beliefs on the
desired behavior. 23
The application of the k-means method
over the data which contained information about 185 stu-
dents in one e-course, at a higher education institution,
resulted with the following three groups:
? Group 0 contained 84 students. ? Group 1 contained 82 students. ? Group 2 contained 19 students.
Figure 2 represents the groups of the students in a form
of a tree, while Figure 3 represents the plot with the move-
ments of the value of the variable “Context” according to
the range of the centroid values.
Figure 2 shows the groups of students in a form of a tree.
According to Table 1, which is a centroid table, group 0
contains the students who had the lowest access to the
content in the e-course. This group shows weekly down-
loading activity of materials from LEs and lectures. Group
1 contains students who had a medium frequency of access
to e-content. They mostly accessed materials from LEs and
lectures. The least accessed set of materials for this group is
related to forums. In group 2, there are 19 students who had
a high frequency of access to materials from AEs, lectures,
and LEs. Figure 3 represents a plot diagram showing the
movement of groups by the value of the variable “Context”
and the range of the centroid values. According to this
analysis, group 0 contains the students with the lowest
frequency of access to the content in the e-course, and
group 2 contains the students with the highest frequency
of access to materials from the e-learning system.
Interpretation of the results obtained by the decision tree technique
After conducting a cluster analysis, which resulted with one
model showing three groups of students, three decision
trees were created based on these groups. Each decision
tree model represents the behavior of one group of the
students. Figure 4 represents the decision tree demonstrat-
ing the behavior of the students from group 0, Figure 5
represents the behavior of the students from group 1, and
finally, Figure 6 represents the decision tree showing the
behavior of the students from group 2. The variable that
gives the highest information gain is the student’s points
from the mid-term exams. The nodes represent the contents
of the e-c

Place your order
(550 words)

Approximate price: $22

Calculate the price of your order

550 words
We'll send you the first draft for approval by September 11, 2018 at 10:52 AM
Total price:
$26
The price is based on these factors:
Academic level
Number of pages
Urgency
Basic features
  • Free title page and bibliography
  • Unlimited revisions
  • Plagiarism-free guarantee
  • Money-back guarantee
  • 24/7 support
On-demand options
  • Writer’s samples
  • Part-by-part delivery
  • Overnight delivery
  • Copies of used sources
  • Expert Proofreading
Paper format
  • 275 words per page
  • 12 pt Arial/Times New Roman
  • Double line spacing
  • Any citation style (APA, MLA, Chicago/Turabian, Harvard)

Our guarantees

Delivering a high-quality product at a reasonable price is not enough anymore.
That’s why we have developed 5 beneficial guarantees that will make your experience with our service enjoyable, easy, and safe.

Money-back guarantee

You have to be 100% sure of the quality of your product to give a money-back guarantee. This describes us perfectly. Make sure that this guarantee is totally transparent.

Read more

Zero-plagiarism guarantee

Each paper is composed from scratch, according to your instructions. It is then checked by our plagiarism-detection software. There is no gap where plagiarism could squeeze in.

Read more

Free-revision policy

Thanks to our free revisions, there is no way for you to be unsatisfied. We will work on your paper until you are completely happy with the result.

Read more

Privacy policy

Your email is safe, as we store it according to international data protection rules. Your bank details are secure, as we use only reliable payment systems.

Read more

Fair-cooperation guarantee

By sending us your money, you buy the service we provide. Check out our terms and conditions if you prefer business talks to be laid out in official language.

Read more
colle writers

Order your essay today and save 30% with the discount code ESSAYSHELP