Copyright © 2018 Authors. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted
use, distribution, and reproduction in any medium, provided the original work is properly cited.
International Journal of Engineering & Technology, 7 (3.15) (2018) 63-67
International Journal of Engineering & Technology
Website: www.sciencepubco.com/index.php/IJET
Research paper
Evaluation of Clustering Methods for Student Learning Style
Based Neuro Linguistic Programming
Marina Yusoff*,1 , Muhammad Najib Bin Fathi2
1Advanced Analytic Engineering Center (AAEC), Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA, Shah
Alam, Selangor, Malaysia.
2Faculty of Computer and Mathematical Sciences, Universiti Teknologi MARA, Shah Alam, Selangor, Malaysia.
*Corresponding author, e-mail: marinay@tmsk.uitm.edu.my
Abstract
Students’ performance is a key point to get a better first impression during a job interview with an employer. However, there are several
factors, which affect students’ performances during their study. One of them is their learning style, which is under Neurolinguistic Pro-
gramming (NLP) approach. Learning style is divided into a few behavioral categories, Visual, Auditory and Kinesthetics (VAK). This
paper addresses the evaluation of clustering methods for the identification of learning style based on system preferences. It starts with the
distribution of questionnaires to acquire the information on the VAK for each student. About 167 respondents in the Faculty of Computer
and Mathematical Science are collected. It is then pre- processed to prepare the data for clustering method evaluations. Three clustering
methods; Simple K-Mean, Hierarchical and Density-Based Spatial Clustering of Applications with Noise are evaluated. The findings
show that Simple K-Mean offers the most accurate prediction. Upon completion, by using the dataset, Simple K-Means technique esti-
mated four clusters that yield the highest accuracy of 74.85 % compared to Hierarchical Clustering, which estimated four clusters and
Density- Based Spatial Clustering of Applications with Noise which estimated three clusters with 52.69% and 61.68 % respectively. The
clustering method demonstrates the capability of categorizing the learning style of students based on three categories; visual, auditory
and kinesthetic. This outcome would be beneficial to lecturers or teachers in university and school with an automatically clustering the
students’ learning style and would assist them in teaching and learning, respectively.
Keywords: Clustering; Hierarchical Clustering; K-Means; Learning Style; Neuro Linguistic Programming
1. Introduction
This The performance of a student is important to result in a good
student in terms of academics or in other activity involvement [1].
Their performance can be affected by a few variables, external or
internal depending on the student’s capabilities of completing their
tasks, learning style, intentions, environment, communication
skills and societies. Everyone varies from the way they learn in
various kinds of environment. Some of them prefer to listen to the
lectures, some of them prefer to write all the notes that have been
given in the slides, while some of them even prefers a lot of imag-
es during their presentation. This learning style has long been
researched by other professionals, as the studies started from a
psychological study on individual differences in the 1960s and the
1970s range [2]. Various tools and theory had been developed by
the researchers to understand how an individual learns.
As the technology is advancing and growing from time to time,
students are more exposed to different method of learning. A good
performance depends on how the students managed their studies,
their approach during their studies. Every learner has their own
different approach of learning according to some factors such as
learning style [3]. Student in university uses different approach
most of the time to study [4]. A suitable learning style is important
to ensure students can achieve a better result. In universities, there
is still a lack of measurements to see the effectiveness of the sys-
tem preference in learning style. It is very hard to identify which
system preferences that influence the student performances.
Learning styles is the way in which individuals perceive and pro-
cess information in learning situations [5]. To the best of our
knowledge, there is a need to develop a system to assist the identi-
fication of student system preferences related to learning style.
Every person uses multiple senses as such visual, taste, sight,
hearing and touch from time to time. For instance, some people
tend to remember things easily by remembering the voice of the
person, some can recognize their face but unable to recall their
names and some people are able to remember the person by the
things they had done together. These senses are very useful during
learning. In teaching and learning, listening to the lectures is easier
for some students to memorize the points, some of them needs to
do something while lecture and some of them system preferences,
people may exhibit a certain behavior or unique characteristic for
themselves. This kind of system preference often explains in Neu-
ro Linguistic Programming (NLP) communication model. Many
people use this idea to rapport, improve the communication skills,
understanding other people behavior and learning style identifica-
tions. VAK is an acronym for Visual, Auditory and Kinesthetic.
Learning styles are methods for system preferences established in
NLP basic communicator because different people learn different-
ly [6]. Therefore, research on learning style based NLP is a prom
ising idea to assist lecturers or teachers to better understanding
student behavior and match to a suitable learning style. This paper
addresses the clustering methods, evaluation to categorize students
learning style based on VAK system preferences.
64
International Journal of Engineering & Technology
2. Neuro Linguistic Programming System
Preferences
As for visual, the person prefers the depiction of diagrams, graphs,
or charts during their study. Symbols also helped them to under-
stand better rather than audio or physical activities. This prefer-
ence is unique as they prefer to use all the images and symbols,
which could have been depicted, by using words or sentences. In a
visual sense, people who prefer this type of learning style learn
best by reading and watching, as they must see it for them to un-
derstand it [7]. They prefer to visual things in every kind of things
they try to understand as if they have a movie camera playing in
their mind while they think. By using this “movie camera”, they
always recall things easily from what they had captured.
The people with this type of learning style prefer information,
which is heard or spoken rather than seeing and visualizing. They
learn best from verbal discussions during classes instead of watch-
ing the slides. They must hear first to learn and understand what
was being explained [7]. A person with this style of learning pre-
fers to talk loud and even talking to oneself to recall things accu-
rately. Unlike visual style of learning, they prefer to speak first
rather than organizing and sorting their ideas beforehand. They are
also a good listener, but this may become a disadvantage to them
as they are easily distracted by sounds. Kinesthetic people are the
one who prefers physical activities involved. They are poor listen-
ers as they prefer to learn by doing practically and have an out-
going personality. These types of people tend to learn best by
doing hands-on activities [7]. By doing so, they can remember
most of the things they had done, unlike visual and auditory per-
son. However, one of the disadvantages of these types of learners
are they are easily distracted or having a hard time paying atten-
tion as they prefer to do things while learning. They connect to the
reality, as they need experiences, practices or simulations. These
also include videos of how to do things, or a demonstration by an
expert or even case studies.
Multimodality is a term used when a person prefers a few learning
styles in any of the combinations such as visual and auditory, or
kinesthetic and auditory, writing and writing and kinesthetic or
such. The person who has multimodality easily adapts with sur-
rounding very well rather than someone with a singular modality
[8]. However, the person with multimodality is uncommon as they
need to be able to multitask while doing something to be able to
learn faster. This type of modalities is divided into two types. First
is someone who can do things with all their learning styles simul-
taneously, or second, the one who can change their learning style
one after another.
3. Clustering Implementation
3.1 Data Acquisition and Pre-processing
A survey of VAK learning style is distributed to students in the
Faculty of Computer and Mathematical Sciences (FSKM),
University Teknologi MARA. About 167 questionnaires were
answered. It is then pre-processed to prepare the data for clus-
tering method evaluations. VAK scoring is calculated for each
respondent. VAK score is transformed, normalized and orga-
nized in CSV format. A normalization technique as such z -
score scaling formula is used to reduce the range of the data
and the difference of the data will not be too large. For cate-
gorical data such as gender, a common technique, which con-
verts the data into a binomial form, is also used. The example
of pre-processed data is shown in Figure 1.
Fig. 1: Sample of pre-processed data
3.2 Clustering Methods
Clustering methods have been highlighted in many research and
applied in many domains [9-13]. In clustering the idea is not to
predict the target class as like classification, it is more ever trying
to group the similar kind of things by considering the most satis-
fied conditions all the items in the same group should be similar
and no two different group items should not be similar [14]. To
group the similar kind of items in clustering, different similarity
measures should be considered. This paper highlights the evalua-
tion of the three most common techniques in clustering; Density-
Based Spatial Clustering of Applications with Noise (DBSCAN),
Hierarchical Clustering and Simple K-Means. These clustering
methods were developed in Phyton and the plotting of the graph
were using Matplotlib library. Figure 2 is a flowchart to demon-
strate the employment of clustering technique steps. Firstly, the
dataset is loaded into the system. Then, a clustering method is
chosen, whether DBSCAN, Simple K-Mean, or Hierarchical clus-
tering technique is used to cluster the data. With the techniques
chosen, the number of clusters is then determined manually, or by
using Elbow Method [15]. The distance of each instance coordi-
nates is calculated with the centroids. This step determines the
cluster of each centroid. Finally, the clusters are plotted into a
graph for visualization.
Fig. 2: Flowchart of Clustering Employment Steps
4. Analysis of Results
The experiments were performed measure the accuracy of the
three methods; DBSCAN, Simple K-Mean and Hierarchical. Ta-
ble 1 shows the number of clusters identified for the three tech-
niques. For Simple K-Mean clustering and Hierarchical clustering,
the number of clusters predicted were both 4 clusters; V, A, K and
Multimodalities (M). M is a mix of more than two of learning
style meanwhile DBSCAN predicted it to have 3 clusters; V, A, K
instead. This is due to the method of calculating the number of
clusters are different for each type of clustering methods.
Gender
Semester
CGPA
V
A
K
Subject
Course
Style
Total
1
6
3.65
4
1
7
ITT575
CS245
K
12
1
6
2.68
2
6
5
ITT575
CS245
A
13
1
6
2.64
6
1
5
ITT575
CS245
V
12
1
6
3.31
7
1
4
ITT575
CS245
V
12
0
6
2.56
7
1
4
ITT575
CS245
V
12
0
6
2.97
3
7
2
ITT575
CS245
A
12
1
6
2.79
5
1
6
ITT575
CS245
K
12
1
6
3.75
7
1
4
ITT575
CS245
V
12
0
6
2.71
5
4
3
ITT575
CS245
V
12
1
6
2.77
4
5
3
ITT575
CS245
A
12
1
6
3.22
6
1
5
ITT575
CS245
V
12
0
6
3.35
9
3
0
ITT575
CS245
V
12
International Journal of Engineering & Technology
65
Table 1: Number of clusters for each technique
Clustering Method
Number of Clusters
Learning Style
DBSCAN
3
V, A, K
Simple K-Means
4
V, A, K, M
Hierarchical
4
V, A, K. M
4.1 Results of Simple K- Means
Figure 3 shows the data cluster for the technique Simple K- Means which
consists of 4 clusters. The data are well represented in Table 3 shows the
exact value of each clustered data. Figure 3 is the clustered data in the
Visual against Cumulative Grade Point Average (CGPA), Auditory against
Cumulative Grade Point Average, Kinesthetic against Cumulative Grade
Point Average and the last one is showing the clustered data and identifica-
tion of Visual for cluster 1, Auditory for cluster 2, Kinesthetic for cluster 3
and Multimodal for cluster 4 for the students in FSKM.
Fig. 3: The cluster for SimpleKMeans
In Table 3, the confusion matrix shows the correctly clustered data
with a total of 167 data. The data are accurately clustered which
the cluster 0, Cluster 1, Cluster 2 and Cluster 3 represent the learn-
ing style of M, A, V and K respectively. As demonstrated in Table
2, there are 46 instances which are incorrectly clustered. 34 of
them were misclustered under Cluster 0, and 12 of them were
misclustered under Cluster 1.
Table 2: Confusion matrix for SimpleKMeans
System
Cluster No
Preferences
0
1
2
3
K
26
0
0
36
A
0
29
0
0
V
8
0
49
0
M
11
12
0
0
4.3 Result of Hierarchical Clustering
Figure 4 shows the data cluster for the technique Hierarchical
Clustering. The data is well represented in Table 4 which shows
the exact value of each clustered data. Figure 4 is the clustered
data for the Visual against Cumulative Grade Point Average, Au-
ditory against Cumulative Grade Point Average, Kinesthetic
against Cumulative Grade Point Average and the last one is show-
ing the clustered data and identification of Visual for cluster 0,
Auditory for cluster 1, Kinesthetic for cluster 2 and Multimodal
for cluster 3 for the students in FSKM by using the Hierarchical
Clustering technique.
Fig. 4: Data cluster of Hierarchical Clustering
From the Table 3, it shows there are 79 instances, which is incor-
rectly clustered. There are 75 misclustered data in Cluster 0 and 4
misclustered data in Cluster 2. The performance of Hierarchical
clustering is not as good as Simple K-Means technique.
Table 3: Confusion matrix for Hierarchical cluster
System
Cluster No
Preferences
0
1
2
3
K
62
0
0
0
A
13
20
14
0
V
57
0
0
0
M
13
0
0
6
In Figure 4, the data is clustered into 3 clusters. This data is well
represented in the Table 4. Many noisy data are represented with
the black spots.
4.3 Result of DBSCAN
Figure 5 shows the data cluster for the technique DBSCAN which
consists of 4 clusters. The data is well represented in Table 5
which shows the exact value of each clustered data. Figure 5 is the
clustered data for the Visual against Cumulative Grade Point Av-
erage, Auditory against Cumulative Grade Point Average, Kines-
thetic against Cumulative Grade Point Average and the last one is
showing the clustered data and identification of Visual for cluster
1, Auditory for cluster 2, Kinesthetic for cluster 3 and Multimodal
for cluster 4 for the students in FSKM by using DBSCAN tech-
nique.
Fig. 5: Data cluster of DBSCAN
Table 4 shows there are 70 instances which is incorrectly clustered
by using DBSCAN technique. There are 38 misclustered data in
Cluster 0, 2 misclustered data in Cluster 1 and 24 misclustered
data in Cluster 2. The performance of DBSCAN is almost likely to
hierarchical clustering.
Table 4:.Confusion matrix for DBSCAN
System
Cluster No
Preferences
0
1
2
K
44
0
18
A
0
0
29
V
27
30
0
M
14
5
6
Table 5 demonstrates the performance of the three clustering
methods in terms of accuracy of correctness in clustering. Simple
K- Means has the highest value which is higher than the accuracy
for DBSCAN and Hierarchical clustering. Simple K- Means,
which has 53.29% accuracy is significantly higher than DBSCAN
and Hierarchical clustering with 61.68% and 52.69% respectively.
This means that the SimpleKMeans technique has good capability
to cluster the dataset according to its correctness in clustering the
learning styles of the students.
66
International Journal of Engineering & Technology
Table 6: Performances of Simple K- Means, DBSCAN and Hierarchical
clustering
Clustering Method
Accuracy
DBSCAN
61.68 %
SimpleKMeans
74.85 %
Hierarchical
52.69 %
4.4 Further Results of Simple K-Means based CGPA
From the results obtained from the 3 techniques of clustering;
Hierarchical Clustering, DBSCAN and SimpleKMeans, Simple K-
Means provides the highest accuracy of 74.85%. Figure 6 demon-
strates the clustered data shows a significant clustered of data
within the range of -1.0 until -0.5. This clustered data means the
students of FSKM’s learning style is not preferably an Auditory
style of learning since the clustered data shows students with
higher CGPA has little value in Auditory learning style.
Fig. 6: CGPA against Auditory graph
Figure 7 shows that the clustered data displays a very dense clus-
tered of data within the range of 0.5 until 1.0. This clustered data
shows the students of FSKM’s learning style has a high value of
visual learning style. From the clustered data, we can predict the
learning style of students are more to visual learning style. This is
a very common for university students since they are more likely
to be exposed to many visual presentations during lectures. It is
one of the effective methods for students to adapt to the learning
environment of university’s life.
Fig. 7: CGPA against Visual graph
Figure 8 shows the clustered data shows a significant value of
clustered data with high density of cluster from the range of 0.5 to
1.0. This shows the students has a mixed of Kinesthetic learning
style with the other 2 learning style. With the mixed of 2 or more
learning style, the students are well adapted to their study envi-
ronment and can utilize all the learning style to achieve a higher
CGPA in their study.
Fig. 8: CGPA against Kinesthetic graph
5. Conclusion
This paper addresses the evaluation of clustering methods for the
identification of learning style based on system preferences. This
project starts with the distribution of questionnaires to acquire the
information on the VAK for each student. Three clustering meth-
ods were compared namely; DBSCAN, Hierarchical Clustering
and Simple K-Means. The use of clustering methods indicates the
capability of clustering learning style can assist educators to de-
termine the learning style of their students in the early semester.
Hence, it will help in improving the performance of students. It is
evident from the details analysis of CGPA with the association of
learning styles, the findings can be seen clearly as an aid to educa-
tors in adapting an appropriate learning style in school or universi-
ty. In addition, the developed system engine based on Simple K-
Means can be a suitable clustering method that can be automati-
cally predicted the students’ learning styles.
Acknowledgement
The authors express a deep appreciation to the Institute of Re-
search and Innovation, Universiti Teknologi MARA a for the
grant of 600-IRMI/DANA 5/3/ARAS (0020/2016) and the Infor-
mation System Department, Faculty of Computer and Mathemati-
cal Sciences, Universiti Teknologi MARA, Shah Alam, Malaysia
for providing essential support and knowledge for the work.
References
[1] Ali, S., Haider, Z., Munir, F., Khan, H., & Ahmed, A. (2013). Fac-
tors Contributing to the Students Academic Performance: A Case
Study of Islamia University Sub-Campus. America Journal of Edu-
cational Research, 1(8), 283-289.
[2] Curry, L. (1987). Integrating concepts of cognitive or learning
style: A review with attention to psychometric standards. Ottawa,
ON: Canadian College of Health Service Executives.
[3] Reid, J. M. (1987). The Learning Style Preferences of ESL Students.
TESOL Quarterly, 21(1),87
[4] Broadbent, J., & Poon, W. L. (2015). Self-regulated learning strate-
gies & academic achievement in online higher education learning
environments: A systematic review. The Internet and Higher Edu-
cation, 27, 1-13.
[5] Brown, H. D. (2000). Principles of language teaching and learning,
(4th ed.). White Plains, NY: Longman.
[6] O'Connor, J., & Seymour, J. (2011). Introducing NLP: Psychologi-
cal skills for understanding and influencing people. Conari Press.
[7] Kanar, C. C. (1995). The confident student. Boston: Houghton Mif-
flin Company.
[8] Krishna, T. S., Babu, A. Y., & Kumar, R. K. (2018). Determination
of Optimal Clusters for a Non-hierarchical Clustering Paradigm K-
Means Algorithm. In Proceedings of International Conference on
Computational Intelligence and Data Engineering (pp. 301-316).
Springer, Singapore.
[9] Ooi, T. H., Ngah, U. K., Khalid, N. E. A., & Venkatachalam, P. A.
(2000). Mammagraphic Calcification Clusters Using The Region
International Journal of Engineering & Technology
67
Growing Technique. In Proceedings of The New Millennium Inter-
national Conference on Pattern Recognition, Image Processing and
Robot Vision (pp. 157-163).
[10] Seman, A., Bakar, Z. A., & Isa, M. N. (2012). An efficient cluster-
ing algorithm for partitioning Y-short tandem repeats data. BMC
research notes, 5(1), 557.
[11] Pindah, W., Nordin, S., Seman, A., & Said, M. S. M. (2015, April).
Review of dimensionality reduction techniques using clustering a l-
gorithm in reconstruction of gene regulatory networks. In Computer,
Communications, and Control Technology (I4CT), 2015 Interna-
tional Conference on (pp. 172-176). IEEE.
[12] Seman, A., Bakar, Z. A., Sapawi, A. M., & Othman, I. R. (2013). A
Medoid-based Method for Clustering Categorical Data. Journal of
Artificial Intelligence, 6(4), 257-265.
[13] Khalid, N. E. A., Yusoff, M., Kamaru-Zaman, E. A., & Kamsani, I.
I. (2014). Multidimensional Data Medical Dataset Using Interactive
Visualization Star Coordinate Technique. Procedia Computer Sci-
ence, 42, 247-254.
[14] Raja, R., & Tien, N. (2018). Exploring multi-modality tools of Neu-
ro-Linguistic Programming (NLP) to facilitate better learning
among Primary School students.
[15] Bajorski, P., & Bajorski, P. (2011). Clustering – Unsupervised
Learning. In Statistics for
Imaging, Optics, and Photonics (pp. 297-327): John Wiley & Sons, Inc.
