Available via license: CC BY 4.0
Content may be subject to copyright.
Page 1/21
Enhancing E-Learning through AI: Advanced
Techniques for Optimizing Student Performance
Rund Fareed Mahafdah
University of Carthage Carthage
Seifeddine Bouallegue
University of Doha for Science and Technology
Ridha Bouallegue
University of Carthage Carthage
Research Article
Keywords: Machine Learning, Deep Learning, Education data, Articial Intelligence, eLearning
Posted Date: August 12th, 2024
DOI: https://doi.org/10.21203/rs.3.rs-4724603/v1
License: This work is licensed under a Creative Commons Attribution 4.0 International License.
Read Full License
Additional Declarations: No competing interests reported.
Page 2/21
Abstract
The integration of Articial Intelligence (AI) into e-learning has transformed conventional educational
approaches, improving the learning process and maximizing student achievement. This study offers a
thorough examination of how articial intelligence (AI) can be utilized to enhance e-learning results by
employing advanced predictive methods and performance optimization strategies. The main goals
consist of creating an AI-based framework to monitor and analyze student interactions, evaluating the
inuence of online learning platforms on student understanding using advanced algorithms, and
determining the most ecient methods for blended learning systems. AI algorithms, known for their
cognitive ability and capacity to learn, adapt, and make decisions, are employed to analyze and forecast
student performance, thereby improving educational quality and outcomes. The practical results
obtained by implementing machine learning and deep learning models, such as Convolutional Neural
Networks (CNN) and Recurrent Neural Networks (RNN), show substantial enhancements in forecasting
different performance metrics. This research highlights the ability of AI to develop adaptable, effective,
and successful e-learning environments, promoting enhanced academic achievement and customized
learning experiences. The ndings demonstrate that CNN outperformed other deep learning and
machine learning algorithms in terms of accuracy during the prediction phase, showcasing the advanced
capabilities of AI in educational contexts.
I. INTRODUCTION
The development and growth of any nation depend entirely upon the education system followed by the
country. There has been a signicant change in the teaching methodology and systems during the past
ve decades, unlike the older age. Many improvements have been observed and identied with various
techniques integrated with the learning methodology. Researchers have proposed various models for
improvement in teaching and learning methods. The old strategy for teaching and learning at the last
level has required upgrades and improvements for a long time [1].
The mindset of the students towards that school and college was expecting a reformation. Many
surveys are done with the view in mind of working towards the improvement of the education system.
Many authors proposed that various modern approaches would help with the required improvement. The
implementation and integration of these modern technical ideas should be done with the help of
technology [2]. The elements of the education system that are expected to be improved are categorized
as follows:
Classroom Teaching: Widely acknowledged as the most conventional pedagogical approach,
classroom teaching has long been a fundamental means of delivering education for successive
generations. This study arms that classroom instruction continues to be one of the most
participatory and ecient means of disseminating knowledge [2].
Mobile Learning: Mobile learning, also known as self-directed learning, is gaining popularity among
young people and has become an important educational method. A signicant number of pupils
Page 3/21
choose this approach because of its adaptability and ease of use [3].
The use of Learning Management Systems (LMSs) with diverse functionalities is becoming
increasingly popular among students and teachers. Several rms offer licensed and open-source
LMS systems, emphasizing the signicance of adequate training for optimal utilization. Research
suggests that the use of Learning Management Systems (LMS) is widespread in higher education,
especially in the Middle East [4].
Virtual Classrooms: This cutting-edge educational approach enables teachers to establish remote
connections with students. It enables easy access to knowledge and learning, promoting interactive
communication between teachers and students through integrated technology.
Blended Learning: Blended learning is the integration of several teaching approaches, such as the
use of gadgets, LMS systems, virtual tutoring, and mobile learning, to produce a full educational
experience [5].
Smart Classroom: Smart classrooms employ technological gadgets and equipment to improve
student involvement and acceptance. Organizations are creating intelligent classroom tools to
promote interaction and sustain student engagement, utilizing smart boards and panels as
signicant educational resources while reducing the need for human interaction [6].
The education industry has seen a substantial transition towards digital learning settings, particularly
due to global events that need remote education solutions. Although there are many e-learning platforms
available, it is still dicult to ensure that students perform at their best owing to problems such as lack
of engagement, limited accessibility, and the need for personalized learning experiences. AI can improve
e-learning by using predictive analytics to detect and assist students who are at risk, creating
personalized learning routes, and implementing real-time feedback systems. Nevertheless, the use of AI
in e-learning is now in its early phases, and its complete capacity to enhance educational results has not
yet been completely achieved. The main issue is the absence of a comprehensive framework that can
properly integrate AI into e-learning systems in order to accurately anticipate and improve student
performance. Existing e-learning systems often neglect to effectively use the vast quantities of data
produced to provide practical insights for instructors and students. Consequently, there is an urgent need
for research and development in AI technologies to generate predictive models that may greatly enhance
student engagement, retention, and overall performance.
This study aims to provide a thorough analysis of the integration of Articial Intelligence to enhance e-
learning outcomes through the prediction and improvement of student performance. The main goals are
to use IoT to track student interactions in Middle Eastern educational environments, assess the inuence
of online learning platforms on student understanding using sophisticated algorithms, and determine the
most effective approaches for blended learning systems. In addition, articial intelligence algorithms are
used to analyze and predict student performance. The questions to be answered through this research
work are:
Page 4/21
RQ1: Is it possible to improve student performance and solve the problem of their lack of participation
accessibility?
RQ2: How to integrate articial intelligence with e-learning systems and improve its performance?
Our research work is structured as follows: Section 1: the related works that show the previous
approaches to deal with this problem are discussed in Section 2. In Section 3, the methodology that was
utilized is outlined in this section. The utilization of the dataset, the application of preprocessing
methods, and the implementation of articial intelligence techniques, specically machine learning and
deep learning, are the three primary steps that comprise this methodology. The results of the proposed
showcase the outcomes of utilizing machine learning algorithms (RF, DT, XGB, KNN) and deep learning
models (CNN, RNN, LSTM, ANN) to forecast ve distinct class labels: Performance Level, Final Grade,
Adaptivity Level, Happy and Sad, and Focused and Unfocused. are discussed in Section 4. Section 5
includes the conclusions reached in this paper.
Finally, Section 6 concludes the paper and reports suggested possible future works.
II. LITERATURE REVIEW
This section describes some previous works that are related to this paper in many sections.
2.1 Distant Learning with Cognitive Skills
There are many university distance learning systems. Some of these systems emphasize adult learning
more than school-level student modules [7]. Some learning mechanisms prioritize theoretical concepts
over pedagogical models [8]. Associative, cognitive, and situational theories have been reviewed in
distant learning situations. The rst is activity-based learning. However, the second emphasizes learner
abilities.
The third one only addresses learning from certain situations. The three should increase many learner
personality byproducts. Improved thinking power, exclusive learning, reasoning, and a positive, real-time
perspective are supposed to determine system eligibility. Some distant learning models emphasize
technology [9]. Many of them target management, development, accessibility, and environmental factors.
2.2 Frameworks Used in Distant E-learning
The success of e-learning platforms is heavily reliant on the motivational factors of both learners and
mentors, as there are numerous platforms available. The mentor's capacity to articulate the system in an
inspiring manner is vital. Multiple authors have presented diverse frameworks throughout history.
Leontidis et al. [10] proposed a learning framework that utilizes multimedia to increase student
motivation. Similarly, the authors in [11] suggested utilizing an objective learning-based approach in
conjunction with constructive mind development. These frameworks have demonstrated numerous
benets based on research.
Page 5/21
The R2D2 framework developed by Bonk et al. [12] specically targeted online learning and catered to
the diverse preferences of students. It utilized virtual learning, audio lessons, video explanations, and
animations to effectively teach complex concepts. Additionally, it encompassed interactive activities that
allowed for the application of problem-solving skills in real-life situations. Although it achieved success,
it was decient in comprehensive data collection methods.
Online learning initially encountered obstacles during its implementation in China. Educators grappled
with unfamiliar learning models while students were inundated with a large amount of information. In
order to tackle these problems, Shen et al. [13] implemented an intelligent framework that utilizes an
automated questioning system to actively involve students and offer them answers. This approach
successfully identied student learning patterns. Data analysis was utilized to enhance learning and
teaching outcomes through the application of diverse data mining techniques.
2.3 Social Media and Sentiment Analysis
Social media is reportedly one of the most popular ways to spread knowledge today. All people, including
businesspeople and government ocials, use social media to connect and express their opinions.
Twitter, Facebook, Instagram, and others are used to nd businesses in various elds using technology.
Manjoo et al. [14] found that social media was used to endorse or oppose election candidates. In Saudi
Arabia, young people prefer social media.
Integration of social media platforms with e-learning can solve many collaborative, communication-
based, and problem-specic activities, according to Li et al. [15]. Student familiarity with media platforms
makes it easy for them to adapt learning methods with social media-based e-learning systems. This type
of social media can mine student behavior and information to identify cognitive improvement.
Sentiment analysis with social media analysis can reveal a person's mindset on specic topics.
Sentiment analysis based on social media posts can be done using many computational algorithms.
Special APIs and natural language processing engines can identify student sentiment and prospects.
Bollen et al. [16] used social media platforms to identify posts, public opinions, blogs, tweets, and other
psychometric instruments to classify them as anger, happiness, depression, confusion, or tension.
Multilingual social media platforms are a major cause of user misinformation. Since the complete
transition from one language to another is not easy, a minimal linguistic computational solution may
allow for variation in results.
Multiple languages make sentimental analysis of mass opinions dicult. Sentiment analysis and social
media give students a powerful platform to learn and improve. Today, the global platform relies on social
media, and youth are greatly impacted by it. This study recommends using such platforms to measure
student sentiment as a model parameter.
2.4 Use of IoT in Learning
Page 6/21
With the rise of IoT devices in industries and homes, a large population now relies on them. Successful
device integration provides anytime learning solutions. Many mature and practical IOT applications are
deployed worldwide. The popularity and availability of cheaper circuits are driving IoT device use. IoT-
enabled environments now have cheaper sensors, actuators, and other devices. Tablets and
smartphones are affordable.
Additionally, Internet access at the ngertips is affordable. Internet access and low device prices make it
easier to integrate IoT devices with e-learning systems. According to Asseo et al. [17], IoT services in e-
learning offer many benets. Remote laboratories are a key use of IoT-enabled services during the
creation of e-learning environments. Urban laboratories in ruble areas that cannot be upgraded can use
remote laboratories to work and learn locally. Due to low student numbers, rural labs are not
economically viable.
Networking and communication technologies enabled this dream by providing more advanced learning
tools at low prices and web-based remote learning laboratories. Only an Internet-connected computer is
needed to practice in these labs. Cloud, IOT, and mobile computing enable more advanced learning
methods. Along with IoT, Web cameras, RFID circuits, Arduino, Raspberry Pie, and actuator devices are
popular in remote learning.
2.5 Activity and analysis of behavior
An understanding of the student's remote environment can be one of the important factors that will be
responsible for identifying the knowledge delivery mechanism. Distance learning systems do not focus
on the behavioral analysis of a student, which is done with the assistance of IoT-enabled devices. The
eligibility to understand. The activity of an individual is one of the most challenging tasks in the distant
learning paradigms. However, once it is achieved, it can be used to deliver the lessons in a very adaptive
situation and a very simple manner for all the scheduled students in a class.
The analysis of the activities that the students are engaged in is yet another technique for the
identication of the work characteristics of the individual. The body language, as well as the movement
of the student, can be identied, observed, and analyzed with the help of various techniques. As per the
study given by [18], Reading the heartbeat of a student with the help of a smartwatch during an online
exam can be one of the important tasks that will provide the student feedback based on his input. At the
time of answering the heartbeat records, he will be able to identify his mindset and personality to
recognize the situation on a macro level. The behavior of the student can be identied with the help of
this reading to propose nervousness as well as stress, which in return can suggest the pattern of
answers given by an individual.
2.6 The Use of Articial Intelligence
Machine learning has become increasingly signicant in recent times, alongside deep learning
techniques, to identify patterns and trends across various domains worldwide. According to [19], basic
Page 7/21
applications can possess signicant capabilities. Using web access and online data processing, the
inputs are analyzed. Suggestions for improvement can be made by integrating various deep learning
algorithms, resulting in a more powerful e-learning system compared to traditional classroom learning.
The utilization of the widely recognized Naïve Bayesian, Support Vector Machines, and the k-nearest
neighbor algorithm is undeniably benecial in obtaining accurate outcomes. A plethora of deep learning
algorithms are primarily dedicated to data mining services, but they are still in the present
circumstances.
III. METHODOLOGY
The methodology that was utilized to achieve the prediction task in this study is outlined in this section.
The utilization of the dataset, the application of preprocessing methods, and the implementation of
articial intelligence techniques, specically machine learning and deep learning, are the three primary
steps that comprise this methodology. Figure1 presents the step-by-step of the AI conducted in this
study.
3.1. Dataset Description and Merging Process
We utilized a dataset that was constructed by merging three publicly available datasets related to
university student performance:
Dataset 1: Student Performance Prediction Dataset
Dataset 2: Students' Adaptability Level in Online Education
Dataset 3: xAPI-Edu-Data
Additionally, we obtained a dataset from the Kaggle website that specically focuses on the academic
accomplishments of college students in the Middle East. This dataset consists of 34 features
encompassing a range of demographic, academic, and behavioral attributes.
The merging process involved organizing the content of each dataset into a single Excel le, ensuring
that each dataset retained its distinct features and instances. Given that each dataset varied in the
number of features and instances, we adopted a systematic approach to integrate them cohesively.
Firstly, Dataset 1, comprising 15 features, was placed in the initial columns of the Excel le. Following
this, Dataset 2 was appended, starting from a new set of columns to avoid overlapping with the rst
dataset’s columns. Finally, Dataset 3 was added, ensuring that its features and instances followed
sequentially from where Dataset 2 ended.
This method of merging resulted in a comprehensive dataset but introduced numerous missing values
due to the differing feature sets across the original datasets. To address these missing values and
maintain consistency, we employed specic imputation techniques. For numerical features, the missing
Page 8/21
values were lled using the mean of the similar values within the respective columns. This approach
ensured that the numerical data remained statistically coherent. For categorical features, we used the
mode of the respective columns, thereby preserving the categorical distributions within the dataset.
By applying these preprocessing steps, we created a unied and robust dataset that facilitated a
consistent and reliable foundation for our AI algorithms to predict and analyze student performance
effectively. This meticulous merging and preprocessing process ensured that the nal dataset was
comprehensive, with minimized biases and inaccuracies, ultimately enhancing the reliability of our
study’s outcomes.
The primary attributes include gender (male or female), age, and institution type (government or non-
government). IT Student (whether the student is currently enrolled in an IT program: Yes/No), etc...
Figure2 depicts the Dataset that specically emphasizes the academic achievements of college
students.
The dataset also contains information on the duration of the class attended, whether the student uses
the Self Learning Management System (LMS) for self-study, the number of resources visited by the
student, the number, the number of discussion posts made, the number of days the student was absent,
the student's class level, and the student's nal grade, etc.
Additionally, it includes the Adaptivity Level, which can be classied as Low, Moderate, or High. It also
incorporates the Emotional State, which can be categorized as either Happy or Sad, as well as the Focus
Status, which can be described as either Focused or Unfocused.
To facilitate predictive modeling, the features in this paper have been classied into ve distinct class
labels: Performance Level (Low, Medium, High), Final Grade (Pass, Fail), Adaptivity Level (Low, Medium,
High), Emotional State (Happy, Sad), and Focus Status (Focused, Unfocused). The analysis and results
sections of this paper will refer to these class labels as Class One, Class Two, Class Three, Class Four,
and Class Five, respectively. This extensive dataset allows for the utilization of sophisticated machine
learning and deep learning algorithms to forecast and improve different facets of student performance.
3.2. Preprocessing methods
Once the dataset was gathered, we utilized various preprocessing techniques, employing advanced
articial intelligence methods, to ensure its suitability for the prediction process. The following
subsections give details of these steps.
These methods are crucial for enhancing the quality and precision of the predictions. The preprocessing
stages encompass the following:
Performing Missing Values Check: This stage entails detecting any instances of missing data in the
dataset. Missing values can have a substantial impact on the performance of machine learning models;
therefore, they were appropriately addressed through imputation or removal.
Page 9/21
To transform categorical features into numerical values, we employed the LabelEncoder technique. This
approach involves replacing each distinct value in the categorical features with a corresponding
numerical value, starting from 0. This transformation is essential because most machine learning
algorithms necessitate numerical input.
Normalization: To guarantee that all features are conned within a consistent range, we implemented the
MinMaxScaler technique. This method rescales the features to a range from 0 to 1, facilitating the
acceleration of gradient descent convergence and guaranteeing that all features are treated with equal
signicance in AI techniques.
After completing the preprocessing, we utilized various Articial Intelligence techniques to forecast
multiple student-related outcomes, such as the Final Grade, adaptive level, emotional state (Happy or
Sad), and focus status (Focused or Unfocused) during lectures. As illustrated in Fig.4, The machine
learning algorithms used are Random Forest (RF), Decision Tree (DT), XGBoost (XGB), and K-Nearest
Neighbors (KNN). The selection of these algorithms was based on their resilience and eciency in
managing diverse datasets. In addition, four deep learning models to improve the accuracy of
predictions are Convolutional Neural Network (CNN), Recurrent Neural Network (RNN), Long Short-Term
Memory (LSTM), and Articial Neural Network (ANN).
A brief description of each algorithm is as follows:
Random Forest (RF)
The Random Forest methodology is a prominent option in machine learning, falling under supervised
learning methods. It is exible and can handle both classication and regression problems. This strategy
leverages the power of ensemble learning, which combines numerous classiers to solve complicated
issues and improve model performance. A Random Forest classier comprises many decision trees
trained on different subsets of the dataset. The classier increases prediction accuracy by combining
predictions from various trees, commonly averaging them. Increasing the number of trees in the forest
improves accuracy while reducing overtting. Essentially, Random Forest combines the strengths of
multiple decision trees to rene prediction accuracy [4].
Decision Tree (DT)
Decision trees are supervised machine-learning approaches used for regression and classication. They
function by learning a series of nested if-else questions based on the data's properties and then
predicting the outcome. The objective is to build a model that predicts the value of a target variable by
applying fundamental decision rules derived from data attributes. The data is separated recursively
based on feature values to boost information gain or minimize impurity at each step. This stage is
continued until a specied point, such as maximum tree depth, or until further splits do not enhance the
model's performance. Decision trees that are not appropriately trimmed may demonstrate overtting.
XGBoost (XGB)
Page 10/21
XGBoost, short for Extreme Gradient Boosting, is an optimized and distributed gradient boosting library
celebrated for its eciency, exibility, and portability. It implements gradient-boosted decision trees
engineered for rapid and high-performance computation. The fundamental process of XGBoost involves
iteratively constructing multiple decision trees and rening them by minimizing a predened loss
function at each step. This iterative enhancement process underpins XGBoost's reputation for speed and
accuracy, making it a preferred tool in machine learning competitions and diverse applications.
K-Nearest Neighbors (KNN)
The K-Nearest Neighbors (KNN) algorithm is a simple and intuitive method used for classication and
regression. It works on the notion that data points with comparable features belong to the same class or
have similar values. When given a new unlabeled data point, KNN selects the k nearest labeled data
points from the training set and utilizes them to create predictions. It selects the most prevalent class
label among its neighbors for classication but computes the average for regression. The KNN algorithm
calculates the distance between the query instance and all training examples, sorts the distances to nd
the top k nearest neighbors, and assigns the class label by majority vote or nds the average of the
closest neighbors for regression.
Convolutional Neural Network (CNN)
Convolutional Neural Networks (CNNs) are a type of neural network specically built to interpret
structured grid data like photographs. CNNs have grown in popularity due to their ability to learn spatial
hierarchies of features from input data in an automated and adaptive manner. CNN’s major components
are convolutional, pooling, and fully linked layers. Convolutional layers apply a series of lters to the input
image, extracting high-level features like edges, textures, and forms. Pooling layers minimize the spatial
dimensions of the data, lowering computational cost and helping to produce representations insensitive
to tiny translations. These high-level characteristics are used by fully connected layers at the network's
end to perform nal classication or regression tasks.
Recurrent Neural Network (RNN)
Recurrent Neural Networks (RNNs) are designed to handle sequential data in which the order of the data
points is critical. RNNs have a hidden state that maintains information about previous items in the
sequence, allowing them to model temporal dependencies. Recent advancements in RNNs have
concentrated on tackling challenges such as vanishing and bursting gradients, which can impede
learning in extended sequences. Techniques include the use of gating mechanisms, which have
broadened the application of RNNs to complicated sequence problems such as speech recognition,
language modeling, and even real-time forecasting in the nancial market.
Long Short-Term Memory (LSTM)
Long Short-Term Memory (LSTM) networks are a subset of RNNs that capture long-term dependencies
in sequence data. LSTMs solve the vanishing gradient problem of standard RNNs by incorporating a
Page 11/21
memory cell that retains its state over long durations. The LSTM design has three gates controlling the
information ow: an input gate, a forget gate, and an output gate. These gates enable the model to retain
important information over extended periods, making LSTMs particularly effective for tasks such as
language modeling, machine translation, and sentiment analysis.
Articial Neural Network (ANN)
Articial Neural Networks (ANNs) comprise layers of interconnected nodes (neurons), with each
connection assigned a weight. ANNs typically consist of an input layer, one or more hidden layers, and an
output layer. Each neuron's output is calculated by applying a weighted sum of its inputs followed by an
activation function such as ReLU (Rectied Linear Unit) or sigmoid. Recent advances in ANNs include
the creation of deeper and more sophisticated architectures such as Deep Residual Networks (ResNets)
and the use of transfer learning, which involves ne-tuning a pre-trained ANN model for a specic job.
These advancements have increased the applicability of ANNs to tasks including image identication,
natural language processing, and complicated pattern recognition.
IV. EXPERIMENTAL RESULTS
This section showcases the outcomes of utilizing machine learning algorithms (RF, DT, XGB, KNN) and
deep learning models (CNN, RNN, LSTM, ANN) to forecast ve distinct class labels: Performance Level,
Final Grade, Adaptivity Level, Happy and Sad, and Focused and Unfocused. The effectiveness of the
models in predicting student outcomes was assessed by evaluating their accuracy values.
Machine Learning Results
Table1 shows the machine learning results for each class label in the dataset mentioned. In class 1, the
DT gave the best accuracy results compared with the others of 95%. In class 2, the RF gave the best
accuracy results compared with the others at 95%. In class 3, the KNN gave the best accuracy results
compared with the others of 88%. In classes 4 and 5, the DT gave the best accuracy results compared
with the others at 97% and 98%, respectively.
Table 1
Machine Learning Results - Accuracy
Class one Class Two Class Three Class Four Class Five
KNN 0.90 0.92 0.88 0.96 0.97
DT 0.95 0.94 0.87 0.97 0.98
RF 0.94 0.95 0.86 0.95 0.96
XGB 0.91 0.94 0.84 0.955 0.965
Page 12/21
Table2 presents the pairwise relationship between each of the two classes in the dataset in the KNN.
Where examining the connections between class labels offers substantial scientic and academic worth
by uncovering the associations and interconnections among various variables in the dataset. This
analysis deepens the comprehension of the interplay between different factors, such as Performance
Level and Adaptivity Level, which can enhance the precision and resilience of predictive models.
Furthermore, it assists in recognizing signicant inuences that impact student achievements, providing
valuable insights for implementing focused educational approaches to address particular requirements.
Pairwise relationship analysis provides a more comprehensive understanding of the intricate dynamics
of educational data, which in turn helps create more impactful learning interventions and improve
student achievement. The highest value is between class 2 and class 2, with 0.901. at the same time, the
lowest value is between class 5 and class 5 with 0.050.
Table 2
KNN Pairwise Relationships
Class One Two Three Four Five
One 0.741 0.789 0.829 0.557 0.544
Two 0.588 0.901 0.238 0.418 0.855
Three 0.260 0.420 0.714 0.973 0.493
Four 0.157 0.630 0.422 0.251 0.874
Five 0.425 0.885 0.484 0.218 0.050
Table3 presents the pairwise relationship between each of the two classes in the dataset in the DT. The
highest value is between class 1 and class 1, with 0.983. At the same time, the lowest value is between
class 5 and class 2, which is 0.003.
Table 3
DT Pairwise Relationships
Class One Two Three Four Five
One 0.983 0.305 0.288 0.762 0.482
Two 0.731 0.790 0.395 0.084 0.906
Three 0.642 0.204 0.612 0.900 0.399
Four 0.529 0.846 0.684 0.677 0.874
Five 0.522 0.003 0.562 0.518 0.233
Table4 presents the pairwise relationship between each of the two classes in the dataset in the RF. The
highest value is between class 1 and class 3, with 0.974. at the same time, the lowest value is between
class 3 and class 2 with 0.015.
Page 13/21
Table 4
RF Pairwise Relationships
Class One Two Three Four Five
One 0.127 0.743 0.974 0.941 0.338
Two 0.194 0.653 0.619 0.585 0.370
Three 0.680 0.015 0.810 0.681 0.256
Four 0.722 0.121 0.773 0.477 0.391
Five 0.905 0.409 0.324 0.884 0.486
Table5 presents the pairwise relationship between each of the two classes in the dataset in the XGB.
The highest value is between class 2 and class 2, with 0.996. at the same time, the lowest value is
between class 2 and class 1 with 0.056.
Table 5
XGB Pairwise Relationships
Class One Two Three Four Five
One 0.548 0.659 0.940 0.301 0.100
Two 0.056 0.996 0.646 0.487 0.803
Three 0.716 0.905 0.250 0.426 0.067
Four 0.194 0.170 0.269 0.286 0.243
Five 0.880 0.574 0.194 0.284 0.419
Deep Learning Results
The outcomes of the machine learning applied to each class label in the dataset that was mentioned are
presented in Table6. The accuracy results that CNN provided in class 1 were the highest of any of the
other classes, coming in at 97%. When compared to the other methods, the LSTM demonstrated the
highest level of accuracy in class 2, with a score of 96%. When compared to the other methods, the
LSTM demonstrated the highest level of accuracy in class 3, with a score of 89%. With an accuracy rate
of 98% and 98.5%, respectively, the RNN demonstrated the highest level of accuracy when compared to
the other methods in classes 4 and 5.
Page 14/21
Table 6
Deep Learning Results - Accuracy
Class / ML One Two Three Four Five
CNN 0.92 0.93 0.84 0.98 0.98
RNN 0.97 0.95 0.82 0.98 0.985
LSTM 0.95 0.96 0.89 0.94 0.90
ANN 0.92 0.91 0.80 0.925 0.91
Table7 presents the pairwise relationship between each of the two classes in the CNN dataset. The
highest value is between class 2 and class 2, with 0.90. At the same time, the lowest value is between
class 3 and class 5, which is 0.159.
Table 7
CNN Pairwise Relationships
Class One Two Three Four Five
One 0.070 0.901 0.802 0.849 0.432
Two 0.324 0.261 0.201 0.787 0.445
Three 0.709 0.476 0.298 0.504 0.159
Four 0.858 0.882 0.276 0.836 0.411
Five 0.485 0.534 0.488 0.787 0.293
Table8 presents the pairwise relationship between each of the two classes in the dataset in the LSTM.
The highest value is between class 1 and class 4, with 0.990. At the same time, the lowest value is
between class 5 and class 4, with 0.077.
Table 8
LSTM Pairwise Relationships
Class One Two Three Four Five
One 0.930 0.884 0.210 0.990 0.978
Two 0.889 0.616 0.755 0.794 0.353
Three 0.500 0.745 0.005 0.108 0.059
Four 0.372 0.190 0.553 0.133 0.204
Five 0.289 0.898 0.203 0.077 0.569
Table9 presents the pairwise relationship between each of the two classes in the dataset in the RNN.
The highest value is between class 5 and class 5, with 0.907. At the same time, the lowest value is
Page 15/21
between class 5 and class 2 with 0.056.
Table 9
RNN Pairwise Relationships
Class One Two Three Four Five
One 0.494 0.571 0.758 0.788 0.738
Two 0.751 0.243 0.091 0.556 0.202
Three 0.872 0.091 0.161 0.597 0.769
Four 0.033 0.105 0.392 0.616 0.514
Five 0.476 0.056 0.072 0.833 0.907
Table10 presents the pairwise relationship between each of the two classes in the dataset in the ANN.
The highest value is between class 2 and class 1, which is 0.996. At the same time, the lowest value is
between class 1 and class 3, with 0.061.
Table 10
ANN Pairwise Relationships
Class One Two Three Four Five
One 0.955 0.142 0.061 0.657 0.779
Two 0.996 0.836 0.505 0.214 0.945
Three 0.616 0.409 0.773 0.578 0.616
Four 0.650 0.155 0.220 0.572 0.649
Five 0.410 0.993 0.700 0.840 0.319
As a result, the Deep learning algorithms outperformed the machine learning algorithms based on
accuracy values in all classes.
-Test for Performance Evaluation
A T-test was performed to compare the performance disparities among different machine learning and
deep learning models. Fourteen comparisons were chosen at random for this analysis. The T-test results
are presented in the table above. The bar chart presented below depicts the T-Statistic values for the
comparisons, which indicate the performance difference between each pair of models. All of the P-
values are greater than or equal to 0.05, indicating that there are no statistically signicant differences
between the performances of any pair of models in this dataset. Consequently, the performance of these
models on the provided dataset is indistinguishable, and selecting any of these models would produce
equivalent outcomes, as shown in Table11
Page 16/21
Table 11
T-Test Performance Evaluation
Model 1 Model 2 T-Statistic P-Value
KNN DT -0.61813 0.553673
KNN RF -0.23905 0.81708
KNN XGB 0.141201 0.891202
KNN CNN -0.12937 0.900259
KNN RNN -0.42477 0.682195
KNN LSTM -0.09035 0.93023
KNN ANN 1.135235 0.289142
DT RF 0.375823 0.716819
DT XGB 0.674008 0.519302
DT CNN 0.373182 0.718707
DT RNN 0.027472 0.978756
DT LSTM 0.587427 0.573119
DT ANN 1.612855 0.145441
RF XGB 0.34493 0.73904
RF CNN 0.063436 0.950976
RF RNN -0.25107 0.80809
RF LSTM 0.174078 0.866129
RF ANN 1.312454 0.225768
XGB CNN -0.23423 0.820688
XGB RNN -0.49768 0.6321
XGB LSTM -0.2267 0.826344
XGB ANN 0.892625 0.398112
CNN RNN -0.27406 0.790982
CNN LSTM 0.068439 0.947116
CNN ANN 1.064115 0.318338
RNN LSTM 0.384185 0.710851
RNN ANN 1.239355 0.250338
Page 17/21
Model 1 Model 2 T-Statistic P-Value
LSTM ANN 1.284025 0.235069
V. CONCLUSION
This study has claried the signicant capacity for integrating AI technologies in e-learning systems. By
creating an AI-driven system, we have successfully monitored and analyzed student interactions in
Middle Eastern educational settings. Additionally, we have utilized advanced AI algorithms to predict and
improve student performance. Our work has shown the substantial benets that can be achieved by
leveraging these technologies in education.
The main goals of this study were effectively accomplished, particularly in addressing Research
Questions 1 (RQ1) and 2 (RQ2). For RQ1, we demonstrated that student performance could be improved,
and issues related to participation and accessibility were effectively addressed. By utilizing AI algorithms
on a comprehensive dataset, we accurately forecasted a range of performance indicators, including nal
grades, adaptivity levels, emotional states, and focus statuses. This predictive capability has directly
contributed to enhancing student engagement and academic outcomes.
For RQ2, the integration of AI with e-learning systems signicantly enhanced their performance. The
machine learning models, namely Random Forest (RF), Decision Tree (DT), XGBoost (XGB), and K-
Nearest Neighbors (KNN), along with deep learning models like Convolutional Neural Network (CNN),
Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), and Articial Neural Network (ANN),
exhibited strong predictive abilities. The ndings demonstrated that deep learning models consistently
surpassed machine learning models in terms of accuracy.
The results emphasize the signicance of incorporating contemporary technological resources to
establish adaptable, streamlined, and successful e-learning environments. Through the utilization of AI,
educators can acquire immediate and accurate information regarding student performance, enabling
them to address any issues promptly and specically. This not only enhances the process of acquiring
knowledge but also promotes enhanced academic performance and individualized learning routes.
III. FUTURE work
Future research will focus on expanding the dataset to include a more diverse range of students from
various regions, enhancing the predictive models' applicability. Incorporating real-time data from
wearable devices can improve the accuracy of predictions regarding students' emotional and cognitive
states.
Moreover, developing an IoT-based model that collects and feeds data to AI systems for Middle Eastern
university students has proven effective in improving educational outcomes. Creating user-friendly tools
and interfaces for teachers is also crucial for maximizing the benets of AI and IoT technologies,
Page 18/21
facilitating their broader acceptance and implementation in educational institutions. This approach will
play a pivotal role in enhancing education.
Declarations
Author Contribution
Dr.seifeddine and Prof Ridha supervised the work
References
1. Djamolovhich, S., Management, B.: Improvement of education system management based on
modern management approaches. 2(4): p. 1–12. (2019)
2. Jabbarova, A.: THE IMPORTANCE OF THE TEACHING METHOD-THEORY AND ITS APPLICATION.
(2020)
3. Lin, X.F., et al.: Chinese undergraduate students' perceptions of mobile learning: Conceptions,
learning proles, and approaches. 35(3): p. 317–333. (2019)
4. Khan, R.A., Qudrat-Ullah, H.: Adoption of LMS in Higher Educational Institutions of the Middle East.
Springer (2020)
5. Alnahdi, A.J.G.J.E., Training: Blended learning in Saudi Arabia-A review. 2(6): p. 1–7. (2019)
. Alaidarous, K., A.A.J.I.J.o.E, I., Madini: Exploring EFL students’ perception in the blended learning
environment in Saudi technical education context. 3(6): p. 69–81. (2016)
7. Leach, K., Walker, S.: Internet-based distance education: Barriers, models, and new research. In
WebNet World Conference on the WWW and Internet. Association for the Advancement of
Computing in Education (AACE). (2000)
. Kukulska-Hulme, A., Traxler, J.: Mobile learning: A handbook for educators and trainers. Psychology
(2005)
9. Aydemir, M., et al.: A theoretical framework on open and distance learning. 174: p. 1750–1757.
(2015)
10. Leontidis, M., Halatsis, C., J.I.J.o., M., Grigoriadou, L.T.: Using an effective multimedia learning
framework for distance learning to motivate the learner effectively. 6(3): p. 223–250. (2011)
11. Passerini, K., Granger, M.J.J.C., Education: A developmental model for distance learning using the
Internet. 34(1): p. 1–15. (2000)
12. Bonk, C.J., Zhang, K.J.D.E.: Introducing the R2D2 model: Online learning for the diverse learners of
this world. 27(2): p. 249–264. (2006)
13. Shen, R., et al.: An open framework for smart and personalized distance learning. In International
Conference on Web-Based Learning. Springer. (2002)
14. Manjoo, F.: Social media’s globe-shaking power. 16. (2016)
Page 19/21
15. Li, X., Ganeshan, K., Xu, G.: The role of social networking sites in e-learning. In 2012, Frontiers in
Education Conference Proceedings. IEEE. (2012)
1. Bollen, J., Mao, H., Pepe, A.: Modeling public mood and emotion: Twitter sentiment and socio-
economic phenomena. In Proceedings of the International AAAI Conference on Web and Social
Media. (2011)
17. Asseo, I., et al.: The Internet of things: Riding the wave in higher education. 51: p. 11–33. (2016)
1. Ericsson, W., Technology: and Internet. [cited Dec. 15, 2016; (2016).
https://www.ericsson.com/res/docs/2016/consumerlab/wearable-technology-and-the-internet-of-
thingsericsson- consumerlab-2016.pdf
19. Hämäläinen, W., Laine, T.: and E.J.D.M.i.E.-l. Sutinen, Data mining in personalizing distance
education courses. 4: p. 157. (2006)
20. Wang, H.: Pattern Classication with Random Decision Forest, 2012 International Conference on
Industrial Control and Electronics Engineering, Xi'an, China, pp. 128–130, (2012).
10.1109/ICICEE.2012.42
21. Yanli Liu, Y., Wang: Jian Zhang Lecture notes in computer science, (2012)
22. Savargiv, M., Masoumi, B., Keyvanpour, M.R.: A New Random Forest Algorithm Based on Learning
Automata. Comput. Intell. Neurosci. 2021(1), 5572781 (2020).
https://doi.org/10.1155/2021/5572781
Figures
Figure 1
Page 20/21
The Methodology
Figure 2
The dataset focuses on the academic accomplishments of college students.
Figure 3
The Steps of the implementation of articial intelligence techniques, specically machine learning and
deep learning
Page 21/21
Figure 4
Classes of Machine Learning and Deep Learning
