IAES International Journal of Artificial Intelligence (IJ-AI)

Vol. 13, No. 2, June 2024, pp. 1195~1205

ISSN: 2252-8938, DOI: 10.11591/ijai.v13.i2.pp1195-1205  1195

Leveraging machine learning techniques for student’s attention

detection: a review

Eng Lye Lim1,3, Raja Kumar Murugesan2,4, Sumathi Balakrishnan2,3

1
School of Diploma and Professional Studies, Taylor’s College, Subamg Jaya, Selangor, Malaysia
2
School of Computer Science, Taylor’s University, Subang Jaya, Selangor Malaysia
3
Digital Health and Innovations Impact Lab, Taylor’s University, Subang Jaya, Selangor, Malaysia
4
Digital Economy and Business Transformation Impact Lab, Taylor’s University, Subang Jaya, Malaysia

Article Info ABSTRACT

Article history: With the advances of the internet and today's innovation, it has become
conceivable to conduct teaching and learning activities remotely through the
Received Jun 26, 2023 online platform. Existing research says that student’s attention state and
Revised Oct 17, 2023 learning result are strongly correlated. However, despite its importance, this
Accepted Nov 6, 2023 can be a challenging task, as students in general taking an online class may
be in a variety of different environments and may be multitasking or
distracted by other factors. This review paper aims to address these
Keywords: challenges by exploring the opportunities offered by machine learning
techniques in attention detection for effective online teaching and learning.
Attention detection By leveraging machine learning algorithms, which can analyze large
Computer vision volumes of data, including eye-tracking, facial expressions, and body
Electroencephalogram EEG movements, we can develop robust models for attention detection in online
Eye gaze tracking learning environments. This paper reviews the challenges specific to online
Machine learning learning, such as students' attention deficits and learning styles, and
Online learning highlights the limitations of current attention detection methods.
Student attention Furthermore, it provides recommendations to advance attention detection
technology, emphasizing the potential of machine learning to enhance
attention detection technology for effective online teaching and learning.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Raja Kumar Murugesan
School of Computer Science, Taylor’s University
Taylor's Road, Subang Jaya, Selangor, 47500, Malaysia
Email: [email protected]

1. INTRODUCTION
In recent years, online learning has gained significant popularity, offering flexible and accessible
educational opportunities to a diverse range of students. Experiencing from COVID-19 pandemic, which is a
past today, many things have changed cross the industry including education in terms of operations and
services. Though students have returned to their respective education institutions, online teaching and
learning is still practiced at least to some extent. However, one of the significant disadvantages is the
student’s attention that can be distracted by the external stimuli such as text messages and noises from the
surrounding environment when attending online classes [1]. Research studies have indicated a positive
association between attention level and academic performance, and poor attention may result in the students
having difficulty following instructions, slow learning, and completing the tasks on time [2]. Hence,
introducing a student’s attention monitoring system during online learning process is crucial for student’s
learning success.

Journal homepage: http://ijai.iaescore.com

1196  ISSN: 2252-8938

One of the challenges faced by educators in online learning environments is the ability to monitor
and assess students' attention levels [3]. Understanding students' attention patterns is crucial for effective
instruction, personalized feedback, and identifying potential learning difficulties. In a traditional classroom,
teachers monitor students' body language or facial expressions to gauge their attentiveness, which can lead to
incorrect conclusions [4]. Therefore, there is a growing interest in applying machine learning techniques to
detect students' attention in real-time. The motivation of the review paper is to address the need of an
effective attention monitoring way in online learning environments. As the demand of online learning
continues to grow, it is important to understand student’s mental state and optimize students' attention levels
for enhanced learning outcomes in online learning context [5].
The purpose of this review paper is to provide an overview of the impact of attention in learning and
the state-of-the-art machine learning algorithms employed for attention detection. We hope to discover the
strengths, limitation, and the potential applications of these attention detection applications by reviewing the
existing research. Furthermore, we also look at different type of data sources that being used in attention
detection, such as eye-gaze data, facial expressions, and electroencephalography (EEG) signals. The
combination of these data sources and machine learning algorithms holds promise for constructing effective
attention detection models.
The objectives of this review paper are: (i) to identify the relationship of learning style and attention
that influence students' attention in online learning (ii) to review the performance of the machine learning
algorithms in terms of accuracy, reliability, and scalability in different attention detection approaches and (iii)
to discuss the implications of attention detection in online learning environments. By reviewing the existing
research, we hope to provide insights into the potential benefits and challenges associated with the use of
machine learning techniques for attention detection in online learning.
The remainder of the paper is organized as follows. Section 2 is the impact attention in learning, that
discuss about the importance of attention in student learning. Section 3, student’s attention detection and
prediction approaches, the finding and discussion is in section 4. Section 5 concludes the paper with some
future recommendations.

2. THE IMPACT OF ATTENTION IN LEARNING

Attention is a fundamental cognitive function and a prominent area of research within the field of
neuroscience. Attention deficits can significantly impact the learning process and various aspects of daily
life. Individuals experiencing attention deficits often exhibit behavioral challenges, including difficulty
following instructions, maintaining focus on tasks, and engaging in effective social interactions [6]. Such
individuals may struggle to stay attentive to their responsibilities and assignments, resulting in reduced task
completion and suboptimal learning outcomes. Concentration difficulties are a primary contributor to
ineffective learning and hinder the successful completion of tasks. Past research has aimed to understand
whether learners are more active during learning tasks or if they can better retain learned content at a later
time [7]. Investing attention during learning processes can be highly beneficial, aiding learners in the more
efficient processing and encoding of information. However, individuals with attention deficits may not
always recognize their attention-related challenges. In response, various protocols and interventions have
been developed to assist students with attention deficits, aiming to enhance their attention, reduce
inattentiveness, and facilitate a more productive learning experience.
Recent studies examined the impact of remote learning practices and difficulties during initial stay-
at-home orders during the COVID-19 pandemic in adolescents with and without attention deficit
hyperactivity disorder (ADHD). Teaching and learning are conducted via online platforms for remote
learning. A total of 238 adolescents (132 males; 118 with ADHD) aged 15.64–17.99 years and their parents
participated in this study. Students were self-rate their experience about remote learning. The result shows
that adolescents with and without ADHD are facing concentrating difficulties while attending remote
learning [8]. As remote learning starts gaining its popularity, the studies for attention detection is important to
help the students to overcome the difficulties to stay attentive during the online learning.
In the education field, research has consistently demonstrated the significant influence of sustained
attention on learners' performance and motivation, especially in the context of online instruction. A study
conducted by Chen and Wang [9], highlighted the potential of attention monitoring through
electroencephalogram (EEG) brain wave signals to enhance individual learners' performance. This approach
offers valuable support to teaching assistants and online instructors, enabling them to boost the sustained
attentiveness of learners, particularly those with lower attention levels during online instructional activities.
The study's findings revealed that learners in the experimental group outperformed their counterparts in the
control group, exhibiting not only better learning performance but also improved sustained attentiveness.
These results affirm the pivotal role of attention monitoring in promoting attentiveness and, consequently,

Int J Artif Intell, Vol. 13, No. 2, June 2024: 1195-1205

Int J Artif Intell ISSN: 2252-8938  1197

enhancing overall learning performance. These studies collectively underscore the profound impact of
learners' attention levels on their learning styles within educational contexts. Understanding students' unique
learning styles not only aids educators in tailoring their instructional approaches but also empowers students
to adapt more effectively to successful learning practices. Given that attention plays a central role in
prioritizing and applying concepts and information, it follows that individuals with attention deficits may
struggle to efficiently process and apply information. Therefore, the level of attention significantly influences
individuals' learning styles, and by enhancing attentiveness, we can aspire to achieve the learning outcomes.

2.1. The relationship of learning style and attention during learning

Attention represents the early phase in the process of learning. We constantly confronted with a
multitude of sensory information or stimuli which is impossible to receive and process all of them [10].
Consequently, we engage in a cognitive process known as selectivity, wherein we choose which stimuli to
focus our attention on. This selective mechanism is orchestrated by the brain's attention control system [11],
which serves the dual purpose of swiftly responding to abrupt environmental stimuli and aiding us in aligning
our attention with our goals and the demands of varying situations. This facet of the brain also plays a crucial
role in regulating alertness and arousal within the reticular activating system (RAS), located in the frontal
lobe. Information that gains priority in our attention is subsequently stored and subjected to the learning
process, giving rise to distinctive learning styles that differ among individuals [12].
Attention is a mental process that enables us to focus our attention on one thing while ignoring other
things. In the classroom, we use various strategies to direct students’ attention to the learning tasks and
activities. These include directing students’ attention toward a specific activity or task by using visual cues
(e.g., pointing out a word in the text), auditory cues (e.g., singing a song) or verbal cues (e.g., telling students
about an upcoming quiz). Hence, various studies have been conducted to understand student’s learning style
in order to facilitate their learning.
A learning style can be understood as an approach that optimizes an individual's learning process
[13]. The extent of a learner's attentiveness significantly influences their learning style. Attention serves as
the gateway through which information is received, and working memory processes this incoming
information, imparting meaning to it [14]. The level of attention, therefore, undeniably impacts a student's
success, as it shapes their learning style. Various frameworks for learning styles have been devised to
examine how learners assimilate information through their selective attention mechanisms. However, each
person possesses a unique preference for particular learning styles, irrespective of their deficits and abilities.
This preference reflects an individual's favored approach to learning. These learning style frameworks
[15]–[17] have found application in the fields of education and health sciences, with a particular emphasis on
distinct learning outcomes, each influenced by the level of attentiveness.
Akaneme et al. [18], conducted a comparative analysis to assess students' learning style preferences,
both with and without attention deficits. The study employed a survey research design in which 158 students
participated, with 73 having attention deficits and 85 without. The research utilized the Fleming VARK
questionnaire and ADHD questionnaire to gather data. The results showed that a majority of students
displayed a preference for the visual learning style. Among students without attention deficits, the preference
leaned toward learning through watching, listening, and reading text materials. In contrast, students with
attention deficits exhibited a preference for hands-on activities, followed by watching and visual
demonstrations. Educators can utilize these questionnaires to tailor their teaching strategies, aligning them
with the specific learning styles of the students, and thereby, enhancing the potential for academic success.
Another recent study conducted by Rawandale et al. [19], it was found that urban students, in the
age group of 14-16 years in Dhule, Maharashtra, India, had a preference for the visual learning style over
kinesthetic and auditory styles. In contrast, rural students within the same age group exhibited a preference
for auditory learning over kinesthetic and visual styles. The study collected data from 50 students each from
rural and urban schools. The study's implications suggest that recognizing students' learning preferences can
empower them to employ effective learning strategies and become more self-motivated, thus maximizing
their learning potential. Furthermore, the study underscores the critical connection between teaching content,
attention, and effective learning. Student learning attention is significantly influenced by the configuration of
the online teaching platform, as well as the teaching content and methods employed [20].
Amid the COVID-19 pandemic, an experiment investigating the impact of various online
instructional styles on students' learning attention levels was conducted [21]. The researcher collected EEG
signals to assess the subjects' attention levels during their engagement in online learning, enabling subsequent
analysis. These attention level signals were then harmonized with Mayer's theories of multimedia technology
for a comprehensive analysis. The study enlisted four students who participated in the experiment, which
involved the utilization of three distinct styles of instructional videos. The results revealed that online course
videos featuring an instructor and content simultaneously on the screen, with the instructor pinpointing the
location of the content while explaining it, were more likely to elicit higher levels of student attention.
Leveraging machine learning techniques for student’s … (Eng Lye Lim)
1198  ISSN: 2252-8938

2.2. Challenges of student’s attention in online learning

Online learning environment has gained its popularity over the years. One of the examples is
massive open online course (MOOC) where it offers learners to acquire new knowledge anytime and
anywhere on the online platforms [22]. The online learning systems are usually web-based applications that
distribute, track, and manage courses over the Internet. The demand of MOOC is still increasing over time
[23]. Despite huge numbers of learners participated on MOOC learning platform, research showed that only
5%-10% of the learners completed their courses on a MOOC platform [24] and facing high dropout rate [25].
One of the main challenges that faced by MOOC platform is the learners loss their attention during the
learning process and bring disastrous impact to learning efficiency [23].
One of the challenges that students experience in both physical and online classes is keeping their
attention on the subject being presented to them. Online learning platforms can be conducted as synchronous
and asynchronous approaches. Instructors use online meeting tool such as Zoom, Microsoft Teams, Cisco
Webex, or Google Meet for synchronous online learning. For the synchronous approach, teachers can
provide immediate feedback to students about their learning. One of the drawback when students have the
option to attend the online classes by using their phone, they tend to being distracted by the text messages
from the phone [26]. In contrast, for asynchronous approach, the online learning contents are uploaded to the
online platform and the students access the learning content at their own pace [22]. The content being
presented to the students in asynchronous approach may be overwhelmed and them to shift their focus
elsewhere and have negative impact on some student’s attention span [27], [28]. However, students find that
short videos (with the length of 3 minutes) are less informative or contain insufficient information for their
learning [29]. Hence, an appropriate length of instructional video is important to maximize the student’s
learning impact.

3. STUDENT’S ATTENTION DETECTION AND PREDICTION APPROACHES

Online learning/e-learning has grown at a fast pace especially due to the global pandemic.
Monitoring student conduct is essential so that teachers may easily identify and rectify inappropriate
behaviour. Student attention monitoring is a process that allows educator to observe and record the amount of
time students are spending on various tasks during online classes. This data can be used to help improve
student learning by giving educators insight into how students are performing in their classes. Various
methods and studies have been conducted by using mobile devices, sensors, and with the aids of machine
learning to helps the educator to understand student’s attention in the class. However, it is still a challenging
task to estimate student’s attention in the various dynamic learning environments [30].
Research studies on students’ attention detection can be found in the literature. To review them, we
propose a taxonomy which is shown in Figure 1. First, we divided the existing attention detection methods
into two main categories-objective and subjective methods, based on the degree of student’s involvement in
the attention detection process. Objective methods use external measurements and observations in data
collection that are quantifiable and less influenced by personal biases. Neurophysiological, oculomotor, and
behavioral data are the example of the data using objective method. In contrast, subjective method relies on
students providing subjective feedback on their own attention levels. Self-reporting and interview are the
examples of subjective method of data collection, as it involves students' personal opinions and perceptions
of their own attention levels. In this review, we are focused into objective methods which are computer
vision, eye gaze tracking, and EEG.

3.1. Computer-vision method

Computer vision is a field of study within artificial intelligence that focuses on enabling computers
to understand and interpret visual information. In the context of student attention detection, computer vision
methods can be employed to analyze various visual cues associated with attention and engagement during
learning. In general, this attention detection systems captured the video streams while the students
participated in a learning activity and then analyze the student’s attention during the learning process. The
common indicators for attention detection using this method are facial expression analysis and posture
tracking [31], [32].
The common tools are webcams or surveillance cameras to detect facial expression of the student
during online learning [33]. Typically, webcams are used to provide a constant and non-intrusive manner of
gathering face photos of the students while they are using mobile devices or personal computers for the
learning activities. The sequence of photos is captured to understand the attention state of the students. Over
the past decade, the application of machine learning and computer vision methods has made tremendous
progress in various areas such as automated assessment, security, image data investigation, identity
verification, and surveillance. Mindoro et al. [34] suggested to apply You Only Look Once, Version 3

Int J Artif Intell, Vol. 13, No. 2, June 2024: 1195-1205

Int J Artif Intell ISSN: 2252-8938  1199

(YOLOv3) algorithms in predicting students attentive and non-attentive state based on the face recognition
during class session. However, this system is sensitive to the student’s facial position because the
attentiveness level of the student is determined by whether the students are looking at the camera. Another
attention detection study conducted for offline classes [20] which also incorporated head pose, audio
components, and class state in analysing student’s attention state. Cameras were installed in the classroom
and the students were aware that the cameras were capturing images for data analysis but were not informed
for attention detection purposes. Environmental sound is recorded to capture the different class state namely
lecturing, practice, interaction, and transcription state. Both data were analysed using deep learning algorithm
and the combination of voice and image information improves the accuracy of the learning attention
recognition.

Figure 1. Type of data collection for attention detection

Another study being done to predict student’s attention level is by using Kinect One sensor [35].
The sensor is used to capture the class activities and then to create a feature set that characterizes a student's
facial and body features. These features are then being used to build the machine learning attention model
includes gaze point, lean back, head displacement, eye closed, face deformation, and mouth open. Seven
classifiers were deployed including decision tree, k-nearest neighbors, and decision trees to predict three
attention states, namely high, medium, and low. Despite the combination of Kinnect One sensor and machine
learning methods providing remarkable results, however, the drawback of this approach is the inter-personal
differences among the individuals that make the model difficult to be generalized.

3.2. Eye-tracking method

Eye tracking is another popular method being used to measure student’s attentiveness in class. This
method measures the movements of the eyes and/or body while students are positively or negatively
interacting with the learning environments in a time-varying manner. By analyzing gaze patterns, fixations,
saccades, and other eye movement metrics, researchers can gain insights into students' attention levels and
their engagement with learning materials. The other attentiveness indicator is the blink rate. The blink rate is
low when the students contact with object of interest. In contrast, high blinking rate associate with fatigue
[36]. A study conducted by Veliyath [37] which focus on eyetrackers to model students attention. Tobii 4c
Eyetracker is used to collect eye-gaze data. The environment setup was in a computer lab where students
either to pay attention to lecture or looking at the computer. Throughout the class session, students were to
self-report their engagement state through a pop-up survey from the computer screen. At the end, a machine
learning model that to determine student’s attention level is built in the combination of eye-gaze data
collected and the self-reported data.

Leveraging machine learning techniques for student’s … (Eng Lye Lim)

1200  ISSN: 2252-8938

The other research investigates the use of eye gaze to predict mind wandering during computer-
based learning. Mind wandering can be considered as an indicator of attention lapse during learning tasks.
The study aimed to determine whether gaze data could be used to predict student’s mind wandering when
interact with computer-based learning and also help the student to re-gain the attention. This research covered
different types of classroom setting (online and offline) to identify student’s mind wandering while
interacting with an Intelligent Tutoring System-GuruTutor. The study achieved success rates of 93% in real-
time wandering detection in classroom for online model [38]. By identifying patterns associated with mind
wandering, educators and researchers can gain insights into students' attention levels and adapt their
instructional approaches accordingly.

3.3. EEG data

With the advancement of EEG detection tools, it has become increasingly feasible and cost-effective
to detect attention by monitoring the variations in attention states. Leveraging attention as a bio-signal of the
brain, a feedback mechanism can be devised, allowing learners to recognize physiological changes occurring
during the learning process. This approach is commonly known as EEG-based Neurofeedback for measuring
attention [39]. Studies have been conducted about children who are attending puberty to study about sleep
disorders causing inattention [40]. The results indicated increased theta waves (θ) associated with attention.
The brainwaves frequency can be divided into five waves, namely delta, theta, alpha, beta, and gamma [41],
[42]. Each type of brain wave represents its mental states, and the value can be varied. When one of the five
types of brain waves in our brain is either overproduced or underproduced, it can cause problems [43].
Everyone’s attentiveness is different to the same learning, and that result the EEG signals of
everyone is vary during the learning process. In a study conducted by George et al. [44], EEG data was
collected using mobile sensors, and features were extracted from raw data. Support vector machine (SVM)
classifier was used to calculate and analyse characteristic features that described students' attentiveness. .
Thorugout the learning process, brain wave signals were adopted for collecting EEG signals of the subjects.
Data processing modules and EEG sensing were used for filtering and preparing the collected data.
Subsequently, the EEG signal features were extracted and organized into categories of attentiveness and
inattentiveness. Although EEG-based attention applications aimed at controlling assistive devices through the
utilization of brain waves, new challenges are emerging such as pre-processing strategies [45], techniques to
reduce the noise [46], and types of electrodes to be used in data collection [47].

4. DISCUSSION
Online learning has the advantages of flexibility that allowing the students at their self-pace. It is
also a great opportunity for university or education provide to reach wider range of students locally and
internationally. Technology plays a vital role in term of planning, designing, and delivering of the online
learning. Apart of student motivation and engagement, this review paper explores different learning style
which can be incorporate into online learning which can contribute to the success of online learning.
Subjective attention detection methods such as self-reporting or human observation mechanism are cost
effective but not entirely reliable because the observers may have misinterpretation on everyone’s behavior
and does not provide real time result [37], [48]. Furthermore, monitoring attentiveness of online students
remain challenging because the absence of physical class engagement and feedback from the instructors.
Besides that, People may be unaware of their attention deficit condition, and therefore, it is difficult to
measure using self-reporting instruments. Whether the students are attentizzve throughout the learning
process or not greatly influences their self-efficacy.
Previous research conducted on the vision-based for objective attention detection method. This
method mainly apply face detection, face recognition, face features, and pose estimation for attention
detection [49]. The accuracy of face recognition is highly affected by maintaining the uniformity and the
quality of input images. This method may be useful because students are most likely having lesser movement
during the online class. With that advantage, a study suggested a student’s attention monitoring and alert
system for online classes. The system applied machine learning algorithms to process the image/video of the
students during attending the online class, and then detect their attention level from attentive, yawning, and
dozing state through face landmark [50]. To enhance the precision of the machine learning data analysis, a
study was undertaken that integrated computer vision and EEG data sources [51]. The study investigated the
correlation of the facial expression and the region of the brain activities.
Implementing computer vision methods for attention detection may require specialized hardware or
software to capture high quality image in the online learning context, which can be costly for the students.
Besides, there’s possible raise of ethical concerns related to privacy and data security concern when the
visual of the student attending the online class is being capture.

Int J Artif Intell, Vol. 13, No. 2, June 2024: 1195-1205

Int J Artif Intell ISSN: 2252-8938  1201

Despite studies showing that eye-tracking is effective in attention detection, however, the key
challenge of this method is proper eye-calibration especial for students who are wearing glasses. In order to
produce a more precision result, the calibration needs to repeat a few times for each participants [48].
Another significant disadvantage of video-based eye tracker is the low resolution of the image capture from
the camera [52] and restricting the movement of the students within the range of eye tracking compound [53].
Hence, wearable eye trackers such as mobile eye trackers and head-mounted system are gaining
popularity and more suitable for the real-world applications [30]. One of the concerns of eye gaze method is
the biometrics identification of the student may be exploited during the tracking process [54].

Table 1. Summary of attention detection methods

Author Objective Method Instrument Analysis/Algorithm Result
Mindoro et To predict Computer Web-camera YOLOv3 The proposed
al. [34] student behavior Vision based on Algorithm model acheived
(attentive or not facial 88.6% accuracy
attentive) based expression
on the face
recognition
during class
session.
Zaletelj and To detect Computer Kinect One Decision Tree, K- The proposed
Košir [35] student’s Vision based on sensor nearest Neighbour model achieved
attention during facial and body (KNN) moderate accuracy
lectures in the features of 75.3%
classroom.
George et To use EEG EEG Data (F7, Participant KNN, SVM, RF obtained the
al.[44] signals for the F3, P7, O1, O2, controlled and Random Forest best result and
Al-Nafjan detection of P8, AF4) engaged in (RF) achieved 96% of
and Aldayel student’s simulated train accuracy
[55] attention during
online classes.
Gupta et al. To detect Computer Web-camera and Deep The proposed
[56] student Vision based on instructional Convolutional system achieved
engagement facial video Neural Network 92.58% of
based on expression, (CNN) model accuracy
multimodal eyes blinking,
information in and head
e-learning movement.
context
Gao and Tan To analyze the EEG data (FT7, Instructional SampEN, The instructional
[21] impact of FT8, T7, T8, videos Statistical video that shows
different styles C5, C6, TP7, Analysis (Kruskal- the appearance of
of online TP8, CP5, CP6, Walli’s test, instructor pointing
political course P7, P8) Mann-Whitney U at the content is
videos on test) more likely to
students' elicit higher levels
attention during of students'
the COVID-19 attention.
pandemic using
EEG
Abate et To monitor Blink, gaze, Web camera, Eye Aspect Ratio The heat-map from
al.[36] student’s and student’s synchronous (EAR), statistical the attention
attention during expression online learning, analysis monitoring
online learning. video recording, application
“attention indicates the
prober” attention level of
application students towards to
material
Veliyath et To predict a Eye tracker Tobii 4c RF, SVM, XGB and SVM
al. [37] student’s Eyetracker, Self- Adaptive Boosting achieved higher
attention using report survey (AdaBoost), and accuracy 78%
eye-tracker questions popup Extreme Gradient
method. from screen for Boosting (XGB)
student to
answer.
Srivastava et To examine the Eye- tracking Tobii Pro X230 Statistical analysis Learners fixated on
al.[57] use of eye- data while eye-tracker, a the same area of
tracking to watching Logitech RGB the video-lecture
understand instructional web camera, and slide.
about learners’ video an Optris PI-400
attention thermal camera.
patterns.

Leveraging machine learning techniques for student’s … (Eng Lye Lim)

1202  ISSN: 2252-8938

With the development of EEG detection tools, it became quite feasible and affordable to measure
attention owing to variations in attention states [58]. Using attention as brain ware bio-signals, a feedback
mechanism can be designed where learners can notice the changes in physiological states happening during
the learning process referred to as EEG-based neurofeedback for measuring attention [39]. The evidence
presented thus far supports the idea that use of EEG for online synchronous classes achieved good result to
detect student’s attention level and maintain the student’s attentiveness during the lesson [9]. One significant
advantage of using EEG in attention detection is the incorporation machine learning in the analysing the EEG
data. Summary from Table 1 has indicated that deep learning models, such as CNN and recurrent neural
networks (RNNs), have proven effective in processing EEG signals to identify patterns associated with
attention levels.
On the other hand, despite these recent findings about the role of EEG in attention detection, this
system also has its constraints. EEG modalities, headsets and data analysis methods are the most prominent
challenges face by EEG-based BCI protocol. The preferred modalities in EEG, namely P300, MI, and steady
state visual evoked potential (SSVEP), face issues in signal processing, especially in identifying methods for
feature extraction and reduction [59]. This is because EEG signals are extremely artefact prone, non-
stationary, and non-linear. For delivering remarkable outcomes, plenty of trials are required to concentrate on
cutting down the period of calibration and effective training approaches [60]. Furthermore, EEG headsets
require either liquid or gels on electrodes that are quite uncomfortable for the user. Dry electrodes can also be
used between the scalp and pad [47]. However, the type of electrode to be used in EEG for attention
measurement is a matter of open debate. It has been observed in previous studies that dry electrodes offer
more noise and artefacts as compared to wet electrodes [61]. Besides that, EEG signals can vary significantly
between individuals [62], and thus causing the attention measurement for each student may be challenging in
online learning settings. Investigation about adaptive algorithms that can learn from individual data and
provide personalized attention estimates may be conducted to overcome this issue.
The growth of machine learning has been significant in recent years, and it is expected to continue
to grow in the future. There are huge opportunities for research in student’s attention detection and prediction
in online learning. It has been observed that current short-term studies such as using small sample size of
participants and in a constraint setup, may not necessarily show subtle changes over time. Machine learning
is a type of artificial intelligence that involves training algorithms to make predictions or decisions based on
data. With the increasing availability of data, the development of more powerful and efficient algorithms
could benefit the development of attention monitoring in the context of online learning. Predicting a student's
learning attention is a complex task, as there are many factors that can affect a student's ability to pay
attention and learn. Some potential factors that could be considered when trying to predict a student's
learning attention include their individual learning style, their current level of engagement with the material,
their motivation and interest in the topic, and any external factors that may be impacting their ability to
concentrate. Additionally, machine learning algorithms could be trained on data about a student's behavior,
such as their test scores and performance on assignments, to predict their performances from student’s
learning attention.

5. CONCLUSION
Online learning can reach to a wider audience due to its flexibility and customized time
management for the participants. There are numerous advantages for the educational institutes to adopt
online education as exclusive or complementary teaching tools. However, online learning also faces various
of challenges such as keeping the students’ attentiveness due to the lack of face-to-face interaction and
potential distractions from the online environment. Educators can help to mitigate these challenges by setting
clear expectations and guidelines for the online behavior, providing regular feedback and support, and
engaging learning environment. Understanding the students learning style also can help students stay
engaged and focused on the online learning environment and reduce the potential distraction. Our finding
also shows that machine learning methods hold significant potential for advancing the field of attention
detection, due to their ability to effectively process and analyze complex data from multiple sources of data
such as EEG, eye-tracking devices, facial expression analysis, and body movement sensors. These techniques
allow for the extraction of meaningful patterns and features that can be used to infer students' attention levels.
By integrating machine learning to the existing attention detection techniques, researchers can develop more
effective strategies to enhance student engagement in online learning, adapt teaching materials to individual
needs, and ultimately, improve learning outcomes. As the field of attention detection continues to evolve, it is
essential for future research to explore the potential synergies between these methods, with the goal of
creating a more holistic understanding of student attention and its role in the learning process. Future
direction of the study can be expanded with multimodal data analysis that involved data from various

Int J Artif Intell, Vol. 13, No. 2, June 2024: 1195-1205

Int J Artif Intell ISSN: 2252-8938  1203

sources, such as EEG, eye-tracking, webcam, and behavioral data, to create comprehensive models for
understanding and predicting students' attention.

ACKNOWLEDGEMENTS
The authors would like to extend their gratitude to the School of Diploma and Professional Studies,
Taylor's College, and Taylor's University, Malaysia, for their unwavering support and encouragement
throughout this research. Special thanks are also due to the reviewers, with acknowledgment of the Chief
Editor, for their valuable feedback.

REFERENCES
[1] O. B. Adedoyin and E. Soykan, “Covid-19 pandemic and online learning: the challenges and opportunities,” Interactive Learning
Environments, vol. 31, no. 2, pp. 863–875, Sep. 2023, doi: 10.1080/10494820.2020.1813180.
[2] B. Çeken and N. Taşkın, “Multimedia learning principles in different learning environments: a systematic review,” Smart
Learning Environments, vol. 9, no. 1, Apr. 2022, doi: 10.1186/s40561-022-00200-2.
[3] S. Ma, T. Zhou, F. Nie, and X. Ma, “Glancee: An adaptable system for instructors to grasp student learning status in synchronous
online classes,” in Conference on Human Factors in Computing Systems - Proceedings, in CHI ’22. ACM, Apr. 2022. doi:
10.1145/3491102.3517482.
[4] Z. J. Zheng, G. J. Liang, H. Bin Luo, and H. C. Yin, “Attention assessment based on multi-view classroom behaviour
recognition,” IET Computer Vision, Sep. 2022, doi: 10.1049/cvi2.12146.
[5] A. Nandi, F. Xhafa, L. Subirats, and S. Fort, “Real-time emotion classification using eeg data stream in e-learning contexts,”
Sensors, vol. 21, no. 5, pp. 1–26, 2021, doi: 10.3390/s21051589.
[6] K. Rubia, “Cognitive neuroscience of attention deficit hyperactivity disorder (ADHD) and its clinical translation,” Frontiers in
Human Neuroscience, vol. 12, 2018, doi: 10.3389/fnhum.2018.00100.
[7] F. Puyjarinet, V. Bégel, R. Lopez, D. Dellacherie, and S. Dalla Bella, “Children and adults with attention-deficit/hyperactivity
disorder cannot move to the beat,” Scientific Reports, vol. 7, no. 1, Sep. 2017, doi: 10.1038/s41598-017-11295-w.
[8] S. P. Becker et al., “Remote learning during COVID-19: examining school practices, service continuation, and difficulties for
adolescents with and without attention-deficit/hyperactivity disorder,” Journal of Adolescent Health, vol. 67, no. 6, pp. 769–777,
2020, doi: 10.1016/j.jadohealth.2020.09.002.
[9] C. M. Chen and J. Y. Wang, “Effects of online synchronous instruction with an attention monitoring and alarm mechanism on
sustained attention and learning performance,” Interactive Learning Environments, vol. 26, no. 4, pp. 427–443, 2018, doi:
10.1080/10494820.2017.1341938.
[10] A. V. Fisher, “Selective sustained attention: a developmental foundation for cognition,” Current Opinion in Psychology, vol. 29,
pp. 248–253, 2019, doi: 10.1016/j.copsyc.2019.06.002.
[11] G. West, D. R. Shanks, and C. Hulme, “Sustained attention, not procedural learning, is a predictor of reading, language and
arithmetic skills in children,” Scientific Studies of Reading, vol. 25, no. 1, pp. 47–63, Apr. 2021, doi:
10.1080/10888438.2020.1750618.
[12] A. S. Jayram, TS and Bouhadjar, Younes and McAvoy, Ryan L and Kornuta, Tomasz and Asseman, Alexis and Rocki, Kamil and
Ozcan, “Learning to remember, forget and ignore using attention control in memory,” arXiv preprint arXiv:1809.11087, vol. 1,
2018, doi: 10.48550/arXiv.1809.11087.
[13] A. R. Knoll, H. Otani, R. L. Skeel, and K. R. Van Horn, “Learning style, judgements of learning, and learning of verbal and visual
information,” British Journal of Psychology, vol. 108, no. 3, pp. 544–563, Sep. 2017, doi: 10.1111/bjop.12214.
[14] D. P. Dirette and M. A. Anderson, “The relationship between learning style preferences and memory strategy use in adults,”
Occupational Therapy in Health Care, vol. 30, no. 3, pp. 245–254, 2016, doi: 10.3109/07380577.2015.1138015.
[15] M. S. Ibrahim and M. Hamada, “Adaptive learning framework,” in 2016 15th International Conference on Information
Technology Based Higher Education and Training, ITHET 2016, IEEE, Sep. 2016, pp. 1–5. doi: 10.1109/ITHET.2016.7760738.
[16] S. Muruganandam and N. Srinivasan, “Appraisal of Felder - Silverman learning style model with discrete data sets,” Indian
Journal of Science and Technology, vol. 9, no. 10, pp. 1–4, 2016, doi: 10.17485/ijst/2016/v9i10/88992.
[17] J. Stander, K. Grimmer, and Y. Brink, “Learning styles of physiotherapists: A systematic scoping review,” BMC Medical
Education, vol. 19, no. 1, pp. 1–9, 2019, doi: 10.1186/s12909-018-1434-5.
[18] U. E. C. Immaculate N. Akaneme, Onah Uchechukwu Hope, “Comparative analysis of learning style preference of students with
and without attention deficit hyperactivity disorder (ADHD),” Journal of the Nigerian Council of Educational Psychologists, vol.
11, no. 1, 2020.
[19] T. Rawandale, S. Achuthan, S. Doss, A. V, and V. B, “Learning style preferences among the urban and rural schoolchildren,”
National Journal of Physiology, Pharmacy and Pharmacology, vol. 10, no. 5, pp. 379–381, 2020, doi:
10.5455/njppp.2020.10.02053202007032020.
[20] X. Ling, J. Yang, J. Liang, H. Zhu, and H. Sun, “A deep-learning based method for analysis of students’ attention in offline
class,” Electronics (Switzerland), vol. 11, no. 17, p. 2663, 2022, doi: 10.3390/electronics11172663.
[21] Q. Gao and Y. Tan, “Impact of different styles of online course videos on students’ attention during the COVID-19 pandemic,”
Frontiers in Public Health, vol. 10, p. 858780, Apr. 2022, doi: 10.3389/fpubh.2022.858780.
[22] B. Kang, “How the COVID-19 pandemic is reshaping the education service,” in The Future of Service Post-COVID-19 Pandemic,
Volume 1, Springer Singapore, 2021, pp. 15–36. doi: 10.1007/978-981-33-4126-5_2.
[23] T. Robal, Y. Zhao, C. Lofi, and C. Hauff, “Webcam-based attention tracking in online learning: A feasibility study,” in
International Conference on Intelligent User Interfaces, Proceedings IUI, in IUI’18. ACM, 2018, pp. 189–197. doi:
10.1145/3172944.3172987.
[24] K. Jordan, “Initial trends in enrolment and completion of massive open online courses,” International Review of Research in Open
and Distance Learning, vol. 15, no. 1, pp. 133–160, 2014, doi: 10.19173/irrodl.v15i1.1651.
[25] M. Rõõm, M. Lepp, and P. Luik, “Dropout time and learners’ performance in computer programming moocs,” Education
Sciences, vol. 11, no. 10, p. 643, 2021, doi: 10.3390/educsci11100643.
[26] J. S. Mendoza, B. C. Pody, S. Lee, M. Kim, and I. M. McDonough, “The effect of cellphones on attention and learning: The
influences of time, distraction, and nomophobia,” Computers in Human Behavior, vol. 86, pp. 52–60, Sep. 2018, doi:
Leveraging machine learning techniques for student’s … (Eng Lye Lim)
1204  ISSN: 2252-8938

10.1016/j.chb.2018.04.027.
[27] M. A. Cicekci and F. Sadik, “Teachers’ and Students’ opinions about Students’ attention problems during the lesson,” Journal of
Education and Learning, vol. 8, no. 6, p. 15, 2019, doi: 10.5539/jel.v8n6p15.
[28] I. Rodríguez Longarela, “Measuring engagement and attention by means of electronic response systems,” SSRN Electronic
Journal, 2020, doi: 10.2139/ssrn.3632766.
[29] A. Manasrah, M. Masoud, and Y. Jaradat, “Short videos, or long videos? A study on the ideal video length in online learning,” in
2021 International Conference on Information Technology, ICIT 2021 - Proceedings, IEEE, 2021, pp. 366–370. doi:
10.1109/ICIT52682.2021.9491115.
[30] Y. Wang, S. Lu, and D. Harter, “Multi-sensor eye-tracking systems and tools for capturing Student attention and understanding
engagement in learning: A review,” IEEE Sensors Journal, vol. 21, no. 20, pp. 22402–22413, 2021, doi:
10.1109/JSEN.2021.3105706.
[31] J. Yue et al., “Recognizing multidimensional engagement of E-Learners based on multi-channel data in E-learning environment,”
IEEE Access, vol. 7, pp. 149554–149567, 2019, doi: 10.1109/ACCESS.2019.2947091.
[32] Z. Zhang, Z. Li, H. Liu, T. Cao, and S. Liu, “Data-driven online learning engagement detection via facial expression and mouse
behavior recognition technology,” Journal of Educational Computing Research, vol. 58, no. 1, pp. 63–86, 2020, doi:
10.1177/0735633119825575.
[33] S. Gupta, P. Kumar, and R. K. Tekchandani, “Facial emotion recognition based real-time learner engagement detection system in
online learning context using deep learning models,” Multimedia Tools and Applications, vol. 82, no. 8, pp. 11365–11394, Sep.
2023, doi: 10.1007/s11042-022-13558-9.
[34] J. N. Mindoro, N. U. Pilueta, Y. D. Austria, L. Lolong Lacatan, and R. M. Dellosa, “Capturing Students’ attention through visible
behavior: a prediction utilizing YOLOv3 approach,” in 2020 11th IEEE Control and System Graduate Research Colloquium,
ICSGRC 2020 - Proceedings, IEEE, 2020, pp. 328–333. doi: 10.1109/ICSGRC49013.2020.9232659.
[35] J. Zaletelj and A. Košir, “Predicting students’ attention in the classroom from Kinect facial and body features,” Eurasip Journal
on Image and Video Processing, vol. 2017, no. 1, pp. 1–12, 2017, doi: 10.1186/s13640-017-0228-8.
[36] A. F. Abate, L. Cascone, M. Nappi, F. Narducci, and I. Passero, “Attention monitoring for synchronous distance learning,” Future
Generation Computer Systems, vol. 125, pp. 774–784, 2021, doi: 10.1016/j.future.2021.07.026.
[37] N. Veliyath, P. De, A. A. Allen, C. B. Hodges, and A. Mitra, “Modeling students’ attention in the classroom using eyetrackers,”
in ACMSE 2019 - Proceedings of the 2019 ACM Southeast Conference, in ACM SE ’19. ACM, Apr. 2019, pp. 2–9. doi:
10.1145/3299815.3314424.
[38] S. Hutt et al., “Automated gaze-based mind wandering detection during computerized learning in classrooms,” User Modeling
and User-Adapted Interaction, vol. 29, no. 4, pp. 821–867, 2019, doi: 10.1007/s11257-019-09228-5.
[39] Y. K. Boroujeni, A. A. Rastegari, and H. Khodadadi, “Diagnosis of attention deficit hyperactivity disorder using non-linear
analysis of the EEG signal,” IET Systems Biology, vol. 13, no. 5, pp. 260–266, 2019, doi: 10.1049/iet-syb.2018.5130.
[40] R. B. Tenenbaum et al., “Response inhibition, response execution, and emotion regulation among children with attention-
deficit/hyperactivity disorder,” Journal of Abnormal Child Psychology, vol. 47, no. 4, pp. 589–603, 2019, doi: 10.1007/s10802-
018-0466-y.
[41] G. A. V. Mendes, “Brain-computer interface games based on consumer-grade electroencephalography devices: systematic review
and controlled experiments,” in Mendes2017BraincomputerIG, 2017.
[42] S. D. You, “Classification of relaxation and concentration mental states with eeg,” Information (Switzerland), vol. 12, no. 5, p.
187, Apr. 2021, doi: 10.3390/info12050187.
[43] W. Wan, X. Cui, Z. Gao, and Z. Gu, “Frontal EEG-based multi-level attention states recognition using dynamical complexity and
extreme gradient boosting,” Frontiers in Human Neuroscience, vol. 15, 2021, doi: 10.3389/fnhum.2021.673955.
[44] F. P. George, I. M. Shaikat, P. S. Ferdawoos Hossain, M. Z. Parvez, and J. Uddin, “Recognition of emotional states using EEG
signals based on time-frequency analysis and SVM classifier,” International Journal of Electrical and Computer Engineering
(IJECE), vol. 9, no. 2, p. 1012, Apr. 2019, doi: 10.11591/ijece.v9i2.pp1012-1020.
[45] N. Padfield, J. Zabalza, H. Zhao, V. Masero, and J. Ren, “EEG-based brain-computer interfaces using motor-imagery: Techniques
and challenges,” Sensors (Switzerland), vol. 19, no. 6, p. 1423, 2019, doi: 10.3390/s19061423.
[46] C. Y. Sai, N. Mokhtar, H. Arof, P. Cumming, and M. Iwahashi, “Automated classification and removal of EEG artifacts with
SVM and wavelet-ICA,” IEEE Journal of Biomedical and Health Informatics, vol. 22, no. 3, pp. 664–670, 2018, doi:
10.1109/JBHI.2017.2723420.
[47] S. Park, C. H. Han, and C. H. Im, “Design of wearable EEG devices specialized for passive brain–computer interface
applications,” Sensors (Switzerland), vol. 20, no. 16, pp. 1–16, 2020, doi: 10.3390/s20164572.
[48] M. A. A. Dewan, M. Murshed, and F. Lin, “Engagement detection in online learning: a review,” Smart Learning Environments,
vol. 6, no. 1, 2019, doi: 10.1186/s40561-018-0080-z.
[49] D. Canedo, A. Trifan, and A. J. R. Neves, “Monitoring students’ attention in a classroom through computer vision,” in
Communications in Computer and Information Science, vol. 887, Springer International Publishing, 2018, pp. 371–378. doi:
10.1007/978-3-319-94779-2_32.
[50] M. L. Roy, D. Malathi, and J. D. D. Jayaseeli, “Students attention monitoring and slert system for online classes using face
landmarks,” in 2021 IEEE 4th International Conference on Computing, Power and Communication Technologies, GUCON 2021,
IEEE, Sep. 2021, pp. 1–6. doi: 10.1109/GUCON50781.2021.9573793.
[51] S. Sung et al., “A study on facial expression change detection using machine learning methods with feature selection technique,”
Mathematics, vol. 9, no. 17, p. 2062, 2021, doi: 10.3390/math9172062.
[52] K. Krafka et al., “Eye tracking for everyone,” in Proceedings of the IEEE Computer Society Conference on Computer Vision and
Pattern Recognition, IEEE, 2016, pp. 2176–2184. doi: 10.1109/CVPR.2016.239.
[53] S. Raina, L. Bernard, B. Taylor, and S. Kaza, “Using eye-tracking to investigate content skipping: A study on learning modules in
cybersecurity,” in IEEE International Conference on Intelligence and Security Informatics: Cybersecurity and Big Data, ISI
2016, IEEE, Sep. 2016, pp. 261–266. doi: 10.1109/ISI.2016.7745486.
[54] J. L. Kröger, O. H. M. Lutz, and F. Müller, “What does your gaze reveal about you? on the privacy implications of eye tracking,”
in IFIP Advances in Information and Communication Technology, vol. 576 LNCS, Springer International Publishing, 2020, pp.
226–241. doi: 10.1007/978-3-030-42504-3_15.
[55] A. Al-Nafjan and M. Aldayel, “Predict Students’ attention in online learning using EEG data,” Sustainability (Switzerland), vol.
14, no. 11, p. 6553, 2022, doi: 10.3390/su14116553.
[56] S. Gupta, P. Kumar, and R. Tekchandani, “A multimodal facial cues based engagement detection system in e-learning context
using deep learning approach,” Multimedia Tools and Applications, vol. 82, no. 18, pp. 28589–28615, 2023, doi: 10.1007/s11042-

Int J Artif Intell, Vol. 13, No. 2, June 2024: 1195-1205

Int J Artif Intell ISSN: 2252-8938  1205

023-14392-3.
[57] N. Srivastava et al., “Are you with me? Measurement of learners’ video-watching attention with eye tracking,” in ACM
International Conference Proceeding Series, in LAK21. ACM, Apr. 2021, pp. 88–98. doi: 10.1145/3448139.3448148.
[58] Z. Mohamed, M. El Halaby, T. Said, D. Shawky, and A. Badawi, “Characterizing focused attention and working memory using
EEG,” Sensors (Switzerland), vol. 18, no. 11, p. 3743, Nov. 2018, doi: 10.3390/s18113743.
[59] P. Milz, P. L. Faber, D. Lehmann, T. Koenig, K. Kochi, and R. D. Pascual-Marqui, “The functional significance of EEG
microstates-Associations with modalities of thinking,” NeuroImage, vol. 125, pp. 643–656, 2016, doi:
10.1016/j.neuroimage.2015.08.023.
[60] T. Bucho et al., “Comparison of visual and auditory modalities for upper-alpha EEG-neurofeedback,” in Proceedings of the
Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS, IEEE, 2019, pp. 5960–5966.
doi: 10.1109/EMBC.2019.8856671.
[61] E. C. Leuthardt et al., “EEG headsets with precise and consistent electrode positioning,” 2022 Accessed: Dec. 30, 2023. [Online].
Available: https://patentimages.storage.googleapis.com/6a/3c/94/478dacb3622ef5/US11241183.pdf
[62] E. Gibson, N. J. Lobaugh, S. Joordens, and A. R. McIntosh, “EEG variability: Task-driven or subject-driven signal of interest?,”
NeuroImage, vol. 252, p. 119034, May 2022, doi: 10.1016/j.neuroimage.2022.119034.

BIOGRAPHIES OF AUTHORS

Eng Lye Lim received his B. Sc. in Computer Science from Universiti Teknologi
Malaysia (UTM) in 2002 and his Master of Computer Science from Universiti Malaya (UM)
in 2009. His research interests include EEG, Machine Learning, Deep Learning, Natural
Language Processing and Adaptive Learning. His current research is in student’s attention
detection using EEG. Lim is currently a Ph. D student in Taylor’s University. He can be
contacted at email: [email protected].

Dr. Raja Kumar Murugesan is an Associate Professor of Computer Science,

and Head of Research for the Faculty of Innovation and Technology at Taylor's University,
Malaysia. He has a PhD in Advanced Computer Networks from the Universiti Sains Malaysia.
His research interests include IPv6, Future Internet, Internet Governance, Computer Networks,
Network Security, IoT, Blockchain, Machine Learning and Deep Learning Algorithms. He is a
Senior Member of the IEEE and IEEE Communications Society, member of the Internet
Society (ISOC), member of the Malaysia Data Science Association and associated with the
IPv6 Forum, Asia Pacific Advanced Network Group (APAN), Internet2, and Malaysia
Network Operator Group (MyNOG) member's community. Talks about IoT, Blockchain, AI,
Machine learning, IPv6, Internet Governance, Digital Transformation in the context of
Industry 4.0 (IR4.0) and Sustainable Development Goals (SDG’s), and Futures Thinking at
various international conferences, forums, and events. Obtained substantial funds locally and
internationally, and actively engages in research, consultancy, and outreach activities in
Advanced Computer Networks. He can be contacted at email:
[email protected].

Sumathi Balakrishnan Sumathi Balakrishnan attained her Master of Software

Engineering from University of Malaya, Malaysia, in 2009. She graduated with a Bachelor of
Computer Science from the University Putra Malaysia. Currently, she has been working with
Taylor’s University, Malaysia as a Senior Lecturer. While teaching she is pursuing her
research for her Ph.D. Her research areas involve activity recognition, depression detection
using EEG, smart home and IoT, and applications for elderly activity monitoring. Many papers
have been published based on her research interest. Her email address is
[email protected].

Leveraging machine learning techniques for student’s … (Eng Lye Lim)