ASSIGNMENT 2 FRONT SHEET

Qualification BTEC Level 5 HND Diploma in Computing

Unit number and title Unit 06: Planning a computing project

Submission date Date Received 1st submission

Re-submission Date Date Received 2nd submission

Student Name Student ID

Class Assessor name

Student declaration

I certify that the assignment submission is entirely my own work and I fully understand the consequences of plagiarism. I
understand that making a false declaration is a form of malpractice.

Student’s signature

Grading grid

P5 P6 P7 M3 M4 D2
❒ Summative Feedback: ❒ Resubmission Feedback:

Grade: Assessor Signature: Date:

IV Signature:
ABSTRACT

In the context that the need for quality assurance at Vietnamese higher education institutions
is receiving attention from relevant parties, finding specific and effective measures to ensure
and improve quality is essential. very urgent. To achieve this, educational data analysis is
considered a quite effective solution. This article is designed to give an overview of big data
in the context of higher education, especially in Vietnam as a recently emerging concern in
interdisciplinary research. The article also clearly identifies the types of data that need to be
collected, stored and analyzed at higher education institutions. Thereby, the article raises
existing issues and suggests initial directions for applying this research field to the practice of
higher education.

3
 TABLE OF CONTENTS

I. INTRODUCTION.................................................................................................................6

II. PROJECT CHARTER.......................................................................................................7

1. Project purpose..............................................................................................................7

2. Objectives of the project...............................................................................................7

III. Devise comprehensive project plans for a chosen scenario, including a work and
resource allocation breakdown using appropriate tools. (P5)..............................................9

1. Overview of the project.................................................................................................9

2. Determine project goals and results..........................................................................10

3. Work Breakdown Structure (WBS)..........................................................................16

1.1. WBS definition...................................................................................................16

1.2. Apply WBS into project....................................................................................19

2. Work timeline:.............................................................................................................19

2.1. Gantt definition.................................................................................................19

2.2. Apply GANTT into the project........................................................................20

IV. Produce comprehensive project plans that effectively consider aims, objectives and
risks/benefits for an identified organization (M3)...............................................................23

V. Justify decisions based on research..................................................................................24

VI. Communicate appropriate project recommendations for technical and nontechnical

audiences.................................................................................................................................33

1. For non-technical audience:..............................................................................................33

2. For technical audience:......................................................................................................34

VII. Discuss accuracy and reliability of the different research methods applied.............36

4
VIII. Assess the extent to which the project recommendations meet the needs of the
identified organisation, including fullysupported rationales for planning decisions made
(M4).........................................................................................................................................41

IX. CONCLUSION:...............................................................................................................42

X. REFERENCES..................................................................................................................43

5
6
I. INTRODUCTION

Currently, distance learning has spread widely around the world, leading to many challenges
in education such as monitoring and coordinating learning. These challenges have attracted
the attention of many researchers to provide high-quality and reliable student monitoring
services. Distance learning has provided an effective educational alternative to traditional
learning in higher education. Lecturers in universities have difficulty understanding students'
emotions and unusual behavior during e-teaching sessions and exams. The purpose of this
research is to use computer vision algorithms and deep learning algorithms to develop a new
system to support lecturers in monitoring and managing students during online lessons and
electronic exams. To achieve the set goal, the system uses software methods, computer vision
algorithms and deep learning algorithms. Semi-structured interviews are also used as
feedback to improve the system. The results show that the system achieves high accuracy in
identifying students in real time, tracking students during online lessons and detecting
cheating. Future work could focus on developing additional tools to support students with
special needs and speech recognition to improve the ability of subsequent facial recognition
systems to detect Cheating on electronic exams in distance learning.

The following sections of this article will delve deeper into specific areas where Big Data
technologies can be applied and explore real-life examples of their successful implementation
in academic institutions..

7
II. PROJECT CHARTER

1. Project purpose.

Apply big data technology in detecting academic fraud and ensuring academic integrity
with clear goals.

 First, big data technology helps detect academic fraud by analyzing large amounts of
data from many different sources. From there, signs and patterns of unorthodox behavior
such as copying, equating answers, and using unauthorized support tools can be identified.

 Second, applying big data technology ensures honesty and fairness in the learning
process by monitoring student learning activities. Thanks to that, results and achievements
are evaluated based on each individual's abilities and efforts without being affected by forms
of academic cheating.

 Third, big data technology also helps prevent academic cheating by analyzing data and
identifying cheating patterns. Effective preventive measures such as developing anti-copy
software and integrity monitoring systems can be applied.

 Fourth, the application of big data technology also improves the quality of education by
analyzing learning data and providing appropriate support to enhance learning efficiency.

 Finally, applying big data technology in detecting academic cheating and ensuring
academic integrity helps build a healthy learning environment, encouraging academic
honesty and ethics from the students and create a foundation for the sustainable development
of education.

2. Objectives of the project

The goal of applying big data technology to detect academic cheating and ensure
academic integrity:

 Detecting academic cheating: An important goal is to use big data technology to detect
academic cheating behaviors such as copying papers, equating answers, using
unauthorized support tools, and other forms of fraud. This helps protect integrity and
fairness in the learning assessment process.

 Building a continuous monitoring system: The goal is to deploy a monitoring system

using big data technology to monitor student activities during the online and offline
8
 learning process. This helps detect fraud early and intervene promptly to prevent and
handle it.

 Analyzing data to detect fraud: The goal is to apply big data analysis techniques such
as data mining, machine learning, and statistics to find potential fraud in learning data.
practice. This helps identify patterns, trends and unorthodox behavioral patterns in
assessment and submission.

 Building an automatic fraud detection system: The goal is to develop an automatic

system based on big data technology to detect academic fraud. The system will use
algorithms and predictive models to identify cheating behaviors, both during test taking
and in submitting assignments.

 Developing handling and intervention mechanisms: The goal is to identify processes

and mechanisms to handle cases of academic fraud. This includes establishing
investigation, review and resolution processes to ensure fairness and compliance in
handling fraud cases.

 Raising awareness and educating the community: The goal is to create educational
activities and raise awareness within the learning community about the importance of
honesty and personal responsibility in ensuring honesty Learning. This may include
organizing training sessions, developing educational programs on academic ethics, and
encouraging the exchange of information and experiences among members of the
learning community.

With the achievement of these goals comes the application of big data technology in
achieving these goals, the application of big data technology in detecting academic fraud and
ensuring academic integrity will contribute to protecting honesty and fairness in the learning
assessment process, creating a healthy and trustworthy environment, while minimizing the
negative impact of academic cheating on students' quality of education.

9
III. Devise comprehensive project plans for a chosen scenario, including a work and
resource allocation breakdown using appropriate tools. (P5)

1. Overview of the project.

Applying big data technology in detecting academic cheating and ensuring academic integrity
is a potential method to monitor and prevent fraudulent acts in education. Big data
technology enables educational institutions to leverage the power of large-scale data analytics
to identify instances of cheating and promote an academic culture.

Cheating is understood to be an ongoing and somewhat vexing problem for academic

institutions as they increasingly take their educational processes online and impose metrics on
teaching evaluation. Currently, there is no specific solution that can solve the problem of
fraud. Due to the importance of this topic, many solutions and methods have been applied to
help lecturers accurately determine students' emotions through algorithms that can detect
emotions, such as face reader and X press. engine . There have been many successful
artificial intelligence algorithms, especially deep learning algorithms, which have proven
popular in the field of computer vision with the help of convolutional neural networks
(CNN). It is commonly used in image recognition and classification.

This unfair practice can occur in any unorganized online exam. A particular type of test is
where the answer cannot be 'googled' or shared via private chats or social networks because
the test solution requires a clear understanding of the subject. Such knowledge may involve a
logical combination of available materials, mathematical calculations with customized inputs,
and customized exercises.

Every student who accesses online course content, including exams, leaves a digital trace.
These traces, in the form of online actions, are often stored by distance education systems.
Our methodology uses these actions to indicate fraudulent behavior. This mark is available
after the exam has ended or can be processed retroactively several years in advance, based on
the student data retention policy set by the educational institution.

To solve these problems, the goal of this research is to use computer vision algorithms and
deep learning algorithms to develop a new system that supports instructors in monitoring and
managing students during class sessions. and online exam. The main focus is on student
facial recognition, classroom coordination, facial emotion recognition, and cheating detection
in online exams. The system detects, measures and outputs facial features of students. The
effectiveness of the facial emotion recognition system of students in exams is evaluated by
10
instructors through interviews. In addition, the system also supports students in verifying
their identity (face authentication). It also includes a fraud detection system that uses facial
recognition, eye tracking and facial movements. It should be noted that the face detection and
recognition process requires high precision in image processing considering its characteristics
in making the system highly effective and reliable. The system also detects the appearance of
multiple faces in a photo or focuses on the appearance of materials used in a photo, such as
paper, books, phones, or hand and finger movements, which can be one of the cheating
mechanisms in the test.

Additionally, the system has a simple interface that provides easy access to each of these
tools. The system was developed based on semi-structured interviews with instructors to
identify the obstacles and challenges they encountered during online or distance learning.
Their views are also taken in and their comments are addressed to the system at all stages of
its development until the system reaches its latest design. Then, after finalizing the system in
its final form, their opinions were given to measure their satisfaction with the system and the
need to adopt such systems in training strategies from distant.

2. Determine project goals and results

 Specific goal:

Using computer vision algorithms and deep learning algorithms to develop a new system to
support lecturers in monitoring and managing students during online lessons and exams to
detect academic cheating

 Related Studies

A detailed explanation of concepts and topics related to this study is provided in this section.
This forms the key background for understanding this study, in addition to relevant
methodological research, as shown in Table 1.

11
Table 1: Overview of the models studied in the exercise and the amount of data used.

 Time-bound

 Phase 1 ( 30/10/23 - 6/11/23): Project Initiation

- Define project objectives and scope

- Identify key stakeholders and their roles

- Establish project team and assign responsibilities

 Phase 2 ( 7/11/23 - 14/11/23 ) : Data Collection

- Identify relevant data sources

- Determine data collection methods and tools

- Develop data collection processes and protocol

 Phase 3 ( 15/11/23 - 24/11/23 ) :Data Integration

- Define data integration requirements and objectives

- Develop data integration workflows

- Implement data integration processes

12
- Ensure data quality and consistency

 Phase 4 ( 27/11/23 - 12/12/23 ) :Data Processing

- Identify data processing requirements

- Implement data preprocessing techniques

- Apply big data analytics algorithms for fraud detection

- Perform statistical analysis and pattern recognition on the processed data

 Phase 5 ( 13/12/23 - 28/12/23 ) :Fraud Detection Models and Algorithms

- Research and select appropriate fraud detection techniques

- Develop and implement fraud detection algorithms

- Test and validate the effectiveness of the detection models

 Phase 6 ( 29/12/23 - 10/1/24 ) :Real-time Monitoring and Alerting

- Design real- time monitoring systems

- Set up automated alerts and notifications for suspicious activities

- Integrate with existing academic platforms for seamless monitoring

 Phase 7 ( 11/1/24 - 23/11/24 ) :Reporting and Visualization

- Develop data visualization dashboards for fraud analysis

- Generate reports on detected fraud patterns and statistics

- Provide interactive visualization s for stakeholders

 Phase 8 ( 24/1/24 - 14/2/24 ) :Intervention and Mitigation Strategies

- Design intervention strategies for handling detected fraud cases

- Develop educational resources and training programs on academic integrity

- Establish policies and guidelines for academic integrity

- Implement support systems for students and faculty

 RESEARCHERS:

Team Leader:

13
Name: Hoang Manh Dung

Role: Team leadership, management and strategic decision making.

Team Supports :

- Name: Tran Thanh Trung

Role: Data Collection

Duties: Dentify relevant data sources .Determine data collection methods and tools. Develop
data collection processes and protocols

- Name: Pham Thi Mai

Role: Big Data Storage Specialist

Duties: Data Integration and Data Processing

- Name: Hoang Van Huy

Role: Data Analyst

Mission: Support in the data analysis process, perform compatibility tests.

 Communication and Group Meeting Schedule:

Time: Weekly on weekends.

Means of Communication: Online meetings via tools such as Zoom.

Duties: Discuss progress, answer questions, and suggest improvements.

 Project Scope:

The scope of the project applying big data technology in detecting academic fraud and
ensuring academic integrity may include the following aspects:

 Define the objectives and requirements: Clearly define the objectives of the system
and the specific requirements it should meet. This includes determining the types of
academic fraud you want to detect, the data sources you will collect, and the desired level
of intervention and prevention.

14
 Data collection and preprocessing: Identify the relevant data sources for monitoring
student activities, such as online learning platforms, assessment results, assignments, and
interaction logs. Collect and preprocess the data to make it suitable for analysis.

 Data analysis and anomaly detection: Apply big data analytics techniques, such as
machine learning and data mining, to analyze the collected data. Develop models and
algorithms that can detect patterns and anomalies indicative of academic fraud or
dishonest behavior. This may involve building predictive models, clustering techniques,
or rule-based systems.

 Real-time monitoring and intervention: Implement a real-time monitoring system that

continuously analyzes student activities and trigger alerts or notifications when
suspicious behavior is detected. This allows for timely intervention by teachers and
educational staff. The system can provide insights and recommendations for appropriate
intervention strategies.

 Integration of additional fraud detection technologies: Consider integrating other

fraud detection technologies, such as biometric authentication, audio recording, or image
analysis, to enhance the system's capabilities. These technologies can provide additional
verification methods during assessments or remote learning sessions.

 Privacy and compliance: Ensure that the system adheres to privacy regulations and
preserves student data. Implement appropriate security measures to protect the
confidentiality and integrity of the collected data.

 User interface and reporting: Develop a user-friendly interface for teachers and
educational staff to access and interpret the system's findings. Provide comprehensive
reports and visualizations that summarize detected anomalies, interventions taken, and
overall system performance.

 Testing and evaluation: Thoroughly test the system using representative datasets and
evaluate its performance in detecting academic fraud. Collect feedback from users and
make necessary improvements based on their input.

 Deliverables:

The results of the project applying big data technology in detecting nano components in
academia and ensuring academic authenticity may include the following components:

 Framework Data Collection: A well-defined framework for collecting and analyzing

15
 unified link data from multiple sources, ensuring data accuracy, privacy, and security.

 Monitoring system: The product will provide a continuous monitoring system to

monitor student activities during online and offline learning. This system will collect and
analyze data from various sources and apply nano and dishonest behavior detection
methods.

 User Interface: The product will provide a user-friendly and intuitive interface for
teachers and education staff. This interface allows them to monitor student activity, view
alerts and notifications, and take protective measures when necessary.

 Alerts and notifications: The product automatically generates alerts and notifications
when it detects academic or scientific dishonesty. This alert will provide detailed
information about the misbehavior and recommend specific interventions to prevent and
treat the problem.

 Data analysis technology: The product will use big data technologies to analyze and
process volumes of learning data. Models and algorithmic analysis will be applied to
detect anomalous behavior and differential pattern actions in the data.

 Target planning: The system detects the existing planning system scale of the education
system, ensuring data flow, compatibility and scalability. This helps prevent fang work
during the testing process and ensures academic integrity.

 Documentation and support guidance: The product will provide documentation

guidance and support to teachers and education personnel on how to use the monitoring
system and implement protective measures. This document will help them understand the
functions and features of the product and apply them effectively in a learning
environment.

The project's deliverables will play an important role in scientific discovery and ensure
academic integrity. It provides tools and methods for educators and educational staff to track,
analyze and print learning activities to ensure integrity and accuracy in the teaching process.

3. Work Breakdown Structure (WBS).

1.1. WBS definition

16
WBS (Work Breakdown Structure) is an important tool in project management, helping to
divide large work into smaller parts for easy management and control. Represented
hierarchical tree, each node in the WBS represents a specific part of the project. The
application of WBS helps project managers have an overview of the scope of work, and helps
to optimize the use of resources, time and budget. Thanks to WBS, project management
becomes more effective, helping to ensure project progress and achieve successful results
according to original goals. (DEVI & REDDY, 2012)

Figure 1-Work breakdown structure

The WBS is a valuable tool for project managers because it helps them to:

 Define the scope of the project: The WBS helps to ensure that all of the workrequired to
complete the project is identified and included.

 Estimate the cost and duration of the project: The WBS can be used to estimate the
resources and time required to complete each work package, which can then be used to
create a project schedule and budget.

 Assign tasks and responsibilities: The WBS can be used to assign tasks and
responsibilities to team members or teams.

 Track progress and identify risks: The WBS can be used to track the progress of the
project and identify any potential risks or problems early on.

17
 The WBS is a living document that should be updated as needed throughout the project.
As the project progresses, the WBS can be used to track progress, identify and manage
risks, and make necessary adjustments to the project plan.

The purpose of a work breakdown structure (WBS) is to help project managers define,
organize, and manage the work required to complete a project. It is a hierarchical
decomposition of the project into smaller, more manageable work packages. The WBS helps
project managers to:

 Define the scope of the project: The WBS helps to ensure that all of the work required to
complete the project is identified and included. This can help to prevent scope creep,
which is when the scope of the project grows uncontrollably, leading to delays and
increased costs.

 Estimate the cost and duration of the project: The WBS can be used to estimate the
resources and time required to complete each work package. This information can then
be used to create a project schedule and budget.

 Assign tasks and responsibilities: The WBS can be used to assign tasks and
responsibilities to team members or teams. This helps to ensure that everyone knows
what they are responsible for and that there is no duplication of effort.

 Track progress and identify risks: The WBS can be used to track the progress of the
project and identify any potential risks or problems early on. This allows project
managers to take corrective action before the problems become too serious.

In addition to these benefits, the WBS can also be used to:

 Improve communication and collaboration: The WBS can be used to communicate the
project scope and plan to all stakeholders. This can help to improve understanding and
alignment, and reduce the risk of misunderstandings and conflict.

 Facilitate change management: The WBS can be used to manage changes to the project
scope, schedule, or budget. By breaking down the project into smaller work packages, it
is easier to identify and assess the impact of changes.

 Provide a basis for project control: The WBS can be used to track the progress of the
project and measure performance against the project plan. This information can be used
to make necessary adjustments to the plan and ensure that the project stays on
track.Overall, the WBS is a valuable tool for project managers of all types of projects. It

18
 can help to improve efficiency, effectiveness, and communication, and increase the
chances of project success.

To create a WBS for your project, you’ll need information from other project management
documents.

Here are six simple steps to create a work breakdown structure:

 1. Define the Project Scope, Goals and Objectives

Your project goals and objectives set the rules for defining your project scope. Your project
scope, team members, goals and objectives should be documented on your project charter.

 2. Identify Project Phases & Control Accounts

The next level down is the project phases: break the larger project scope statement into a
series of phases that will take it from conception to completion. You can also create control
accounts, which are task categories for different work areas you want to keep track of.

 3. List Your Project Deliverables

What are your project deliverables? List them all and note the work needed for those project
deliverables to be deemed successfully delivered (sub-deliverables, work packages,
resources, participants, etc.)

 4. Set WBS Levels

The WBS levels are what make a work breakdown structure a “hierarchical deconstruction of
your project scope”, as defined by the project management institute in its project management
body of knowledge book (PMBOK). You’ll need to start at the final project deliverable and
think about all the deliverables and work packages needed to get there from the start.

 5. Create Work Packages

Take your deliverables from above and break them down into every single task and sub task
that is necessary to deliver them. Group those into work packages.

 6. Choose Task Owners

With the tasks now laid out, assign them to your project team. Give each team member the
work management tools, resources and authority they need to get the job done.

1.2. Apply WBS into project.

19
As project manager, I created a Work Breakdown Structure (WBS). The WBS breaks the
project into smaller, more manageable tasks and provides a clear structure for project
planning and analysis. Here is the professional WBS for the project:

2. Work timeline:

I will use gantt chart software to create a Work timeline for my project.

2.1. Gantt definition

A project management tool called Gantt is used to visualize project timetables as bar charts.
The start and finish dates of the project's tasks, as well as their relationships, are shown using
Gantt charts. This project management tool, created by Henry Gantt in the 1910s, has grown
in popularity because it makes it simpler to plan, monitor progress, and modify tasks. Gantt
charts offer a high- level picture of the project process and are an effective tool for making
sure projects are finished quickly and effectively. (Wilson, 2003)

20
Horizontal bars that indicate individual jobs or activities are often aligned along the vertical
axis that depicts the time frame in the chart. The task's duration is represented by the length
of each bar, and the bars are arranged according to the start and finish dates of the task.
Linking lines connecting the bars are frequently used to highlight dependencies between
tasks, demonstrating the order in which they must be done.

Figure 3-Gantt chart

Project managers may more effectively distribute resources and team members can better
grasp their responsibilities and deadlines thanks to Gantt charts, which provide a visual
approach to convey project timelines. They are an effective tool for organizing, managing,
and keeping track of projects of all sizes and levels of complexity. Project stakeholders can
monitor project progress and make any necessary adjustments or revisions in response to
changes in the project's needs or time frame with the aid of Gantt charts.

2.2. Apply GANTT into the project.

21
22
23
IV. Produce comprehensive project plans that effectively consider aims, objectives and
risks/benefits for an identified organization (M3)

Applying big data technology in detecting academic fraud and ensuring academic integrity
can bring many important benefits. By leveraging the power of large-scale data analytics,
educational institutions can identify instances of dishonest behavior and promote a culture of
honesty in education. Here are some methods and benefits of applying big data technology in
this field:

 Detecting patterns and anomalies: Big data technology enables mass analysis of learning
data to identify patterns and anomalies. Data analytics algorithms can identify unreliable
learning patterns, such as students with sudden changes in performance or inconsistent
learning patterns. This helps identify potential cases of cheating or academic dishonesty.

 Identify signs of cheating: Big data technology allows educational institutions to collect
and analyze large amounts of learning-related data, including assignments, tests and
other activities. By applying advanced analytics techniques, such as data mining,
machine learning and artificial intelligence, it is possible to identify signs of academic
cheating. For example, finding similar work, duplicate answers, or unnatural learning
patterns can be indicators of academic dishonesty.

 Building a continuous monitoring and warning system: Big data technology allows
building an automatic warning and continuous monitoring system to detect academic
fraud. By automatically identifying and tracking patterns of dishonest behavior, the
system can notify and provide alerts to education staff to conduct investigations and
handle cases of fraud.

 Create a culture of honesty in education: Applying big data technology can also promote
a culture of honesty in education. By using learning data to assess and analyze learning,
educational institutions can create transparency and fairness in student assessment, while
encouraging honesty and ethics. virtue in learning.

However, it should be noted that the application of big data technology in academic
nanodiscovery also creates some risky formulations. These risk errors include:

 Privacy and data security: When collecting and using learning data, it is necessary to
have a definition and policy regarding data privacy and security. Ensure that academic
data is protected and used only for scientific discovery and academic accuracy.
24
 Evaluate errors and depend on data: The biggest technology is subject to errors and
depends on the quality and reliability of data. The discovery of nanotechnology in
academia based on data analysis can be erroneous or inaccurate if the data is not of
sufficient quality or does not represent all aspects of the research.

 Risk of evaluation failure: Using big data technology to discover scientific technology
in academia may face the risk of evaluation failure. There is a need to ensure that
algorithms and data analyzes are designed according to their use and do not introduce
bias or disadvantage to panel biology tools.

 The importance of educational and ethical measures: Although big data technology
can be a useful tool in detecting hardship and ensuring academic integrity, it is only one
part in this process. To perfect the culture of honesty in education, it is necessary to
combine the application of technology with educational and ethical measures.
Educational institutions build environments of trust, create clear policies and regulations,
and encourage academic honesty and ethics.

In summary, applying big data technology to discover academic breakthroughs and ensure
academic accuracy brings many benefits. However, it is necessary to consider and consider
factors such as privacy, data security, assessment errors and reliance on data, the risk of
unfair assessment and the importance of education and ethics to achieve the best results in
employment providing a culture of honesty in education.

V. Justify decisions based on research

In this section, the general methodology used to develop this system with all its tasks is
explained, in addition to adopting the Agile methodology. Firstly, to begin to analyze the
system requirements, the opinions of lecturers and experts were taken to identify the
problems and challenges that they face in distance learning. Secondly, the contents were
designed and drafted, and the experts’ opinions were taken again about the system objectives
and contents. Thirdly, development occurred in several stages, and the designers returned at
each stage to take opinions and comments to improve the development and design of the
system. Fourthly, after the design and development of the system was completed, the experts
tested all system phases. Finally, the system deployment and feedback tracking phase
involved taking the experts’ opinions again about the system tasks. A semi-structured
interview was conducted with lecturers and experts to verify their satisfaction with and need
for this system and to reveal the effectiveness and accuracy of this system. This section also

25
discusses how to collect data, algorithms, and tools used to identify the student’s face and
feelings and then to detect cheating on an e-test. It is worth noting that the Agile
methodology was adopted because it is characterized by its ease of change and easy
modifications. It also saves time and effort by not requiring developers to completely rework
any changes. This system employs software methods, computer vision algorithms, and deep
learning algorithms (i.e., convolutional neural networks) to develop a new system that gives
lecturers the ability to properly coordinate a classroom and manage communication with their
students during the lesson, in addition to ensuring their attention and studying their
behavioral state in classes. Furthermore, many tools were presented to verify identity using
facial feature extraction, as well as a system for detecting cheating by using face detection
and gaze tracking. Moreover, this system has a simple interface that provides easy access to
each of these tools. Figure shows the system development process by an using the Agile
method.

Figure 2. Agile development process.

1. Computer Vision Algorithms

26
A high-level language-written computer system was utilized to build and implement a
computer vision algorithm. The OpenCV Library is a robust source of computer vision
algorithms. Presently coded in C++, OpenCV is free and open-source software. With such a
wide variety of embedded vision processors (OpenCV), the algorithm analysis in this study
focused on Agile to maximize pixel-level processing within the system. Within the Vision
System Toolbox, Agile functions and Simulink blocks were utilized to implement computer
vision algorithms, and they also allowed us to build our own library of functions that were
tailored to the Agile programmable architecture. The functions applied in computer vision
system include:

• Image acquisition: Numerous image sensors from light-responsive cameras work together
to create digital images. The image data produced were an image series. The pixel values
represent the intensity of light in one or more spectral bands.

• Pre-processing: In order to ensure that the data meet the rules presented by the computer
vision approach, the data had to be processed before the method could be applied to image
data in order to extract the required information. To ensure that the image organization
system was accurate, resampling was carried out. To ensure that sensor noise did not
contribute to erroneous information, noise cutback was carried out. To ensure that important
information could be detected, contrast augmentation was conducted. To improve visual
formations at proximate relevant scales, a scale space representation was created.

• Feature extraction: From the data, image characteristics of varying degrees of complexity
were recovered. These consisted of ridges, lines, and edges. We then performed localized
interest locations, such as blobs.

• Detection/Segmentation: During the processing, a decision was made regarding which

image points or areas should be processed further by dividing an image into segments, each
of which included a different object of interest. Image partition into nested scene architecture
was carried out that included the foreground, object groups, individual objects, or visual
salience. Subsequently, we created a foreground match for each frame of one or more videos,
while maintaining the continuity of the videos’ temporal semantics.

• High-level processing: At this stage, the input often consists of a small amount of data,
such as a collection of points or an area of an image that is supposed to contain a certain
object. The data’s compliance with model-based and application-based rules was checked in
this section. The estimation of factors particular to a given object size was performed. Image

27
recognition categorizes captured objects into many groups. Image registration compares and
combines two images of the same object from different angles.

• Decision making: The final decision is made by applying automated inspection programs
via a Pass/Fail system, with a match or no-match in the recognition of images.

2. Convolutional Neural Network (CNN)

CNN as a deep learning system was used to take in an input image, rank various features and
objects within the image, and distinguish between them. In this study, the [70] method of
Mini Xception CNN analysis was utilized to statistically predict facial emotions. One section
of the article describes how CNN is used to identify facial expressions of emotion. CNN
analyzes real-time video frames to forecast the likelihood that each of the seven fundamental
emotional states will occur. The analysis model uses real-time input from the CNN model’s
output data to predict future emotions in students. The two-phase hierarchical system makes
it easier to examine human behavior and make predictions about the future. The seven
fundamental facial emotions of the recognized face were statistically analyzed as the next
phase after face detection. In the second and most important section of this study, which deals
with emotion prediction, the derived data were used as input for the real-time model. A csv
file containing the computed data was sent as the input. The model estimates the percentage
of each emotion for the expected future using the data gathered about a student through time
and a separate set of inputs for each emotion.

3. Data Collection and Tools

n this study, several datasets were used. The first set of labeled faces in the raw data used to
train face verification models was obtained from images available on the Internet, and this
database contains approximately 13,500 images

The second set is the dataset used to train emotion recognition models. These data were taken
from the FER dataset database on Kaggle.com, which includes around 32,300 images of faces
with different facial expressions. The data clarify seven major emotions at 48 × 48 pixels in
grayscale.

The third dataset was used to train the gaze tracking model. This dataset was obtained from
gaze interactions for everybody (GI4E) and consists of 1380 images with a resolution of 600
× 800 pixels. The fourth dataset was used to train the head movement model. The dataset was
obtained from the FEI Face Database and consists of 2800 images of 200 people at 640 × 480
28
pixels. The fifth dataset was used to detect objects depicted in the camera. The dataset was
obtained from common objects in context (COCO) and consists of 80 object categories.

3.4. Combining the Models to Develop the System

The approach was used to merge the various models in order to analyze the datasets. The
method represented the surfaces of images in the class using the potent statistical tools of 3D
morphable models. Additionally, it merges two or more 3DMMs that are constructed from
various datasets and distinct templates, some of which may only partially intersect. The tools
combine two or more models by filling in the gaps in one model’s data with data from the
other using a regressor. The Gaussian Process framework is commonly used to combine
covariance matrices from other models. This is done by creating a new face model that
includes the variability and facial characteristics.

3.5. System Structure

This system was developed on a laptop running Windows 10 using Python 3.3.8. In addition,
this system was developed by dividing it into three systems to achieve its desired goals and to
meet the needs and suggestions of lecturers. Dividing it in this way makes it easier for us to
develop and modify any system and reduces the processing effort for the computer. In this
section, we present the architecture of systems, models used, and the algorithms. Figures 3
and 4 show the user interface and the system stages, respectively.

29
Figure 3. The user interface.

30
 Figure 4. The developed system structure.

5.1. The Student Identification System

Referring to this system, students who attend the class or an examination will be identified
remotely to verify whether the person who attends the lesson or the exam is the same as the
account holder to avoid cases of cheating by identity theft. Using such a system, the stored
images of students must be compared to the images received via the web camera. As a result,
it employs the deep metric learning algorithm and the ResNet-34 algorithm in two stages.
The first stage: facial feature extraction from the images in the database. The second stage is
comparing the faces received via the webcam with the faces stored or existing in the
database. In the beginning, the pictures of the faces that must be recognized are entered,
which are considered the photos of the class students. Therefore, the face is recognized, for
instance, by detecting the presence or even the absence of faces and discovering their
location. Then, about 128 facial features are extracted through an algorithm in the form of
numerical coefficients and stored in a beam. “Deep metric learning” is an algorithm that
depends on adopting an output represented by a beam of integer values instead of the usual
tool with deep learning that accepts a single image as input and where the output is a
classification or description of the picture.

31
Then, the resulting vector from 128 measures is compared with the vectors stored in the
database, and if no match is made with any vector, the face is considered unknown, but if the
vector is matched with an existing vector in the database, it will draw a square around the
student’s image, writing the student’s name as it is saved in the database. The architecture of
this facial recognition network depends on applying the 34-ResNet algorithm. The network
was trained on a (LFW) dataset that contains around 13,250 images belonging to 5794
people, split into 70% of the training data and 30% of the test data

5.2. Interviews

Conducting interviews is one of the most popular methods for gathering data in qualitative
research. Semi-structured interviews were utilized in this study to examine more about the
specifications for a facial recognition system that will detect exam cheating by students as
well as to acquire more insight on how effective the system is for online proctoring and
distance learning. A pre-determined list of questions was used in the semi-structured
interviews (Table 1), although the interviewers (lecturers) were also requested to clarify any
points and go into further detail about some topics. The benefit of semi-structured interviews
is that the researcher has complete control over the data collection process. The setting for
interviews is voluntarily, seeking the interviewers’ consent and informing them about the
confidentiality and anonymity of their participation. There were two forms of interviews
conducted in this study: one was for the lecturers, and the other for the students. Both
interviews were conducted to measure the lecturers’ and students’ perceptions or feedback
towards the effectiveness of the system for facial identification, classroom coordination, face
emotion recognition, and detecting cheating in online exams.

This system was created after interviewing the lecturers (Table 2) on their perceptions of
facial recognition system usage in their teaching and learning processes, particularly in exams
in online or distance learning. These responses and the identified problems were used to
develop the systems. The responses included the difficulties and barriers they encounter and
the efficiency level of the system in recognizing and detecting the students’ cheating
behaviors and emotional expressions. At every level of the system’s development, up until
the system reached the final stage after a 14-week period, the lecturers’ opinions were taken
into consideration. Overall, six university lecturers at Cyprus International University were
interviewed using semi-structured interviewing techniques, in addition to the twenty students
who were chosen at random from a population of 100 undergraduate students at Cyprus

32
International University, from the Department of Management Information Systems in
Nicosia. A simple sampling technique was employed. Then, after the system was completed,
the lecturers’ feedback was obtained to determine how effective and satisfied they were with
it and whether it should be included in distance learning or online learning sessions.

Table 2. Interview questions used.

5.4. Follow-Up

System during the Exam One of the most prominent issues and limitations that face distance
learning is e-exams, as institutions resort to various methods to implement exams and ensure
that there are no cases of cheating. Therefore, in this section, we developed a new system to
follow up the learners during the exam by warning them if they look outside the permitted
exam space on the computer screen or exhibit abnormal behavior. The principle of work was
based on four steps: the first was to verify the face to ensure that the person is present in the
exam using a trained model based on the deep metric learning algorithm and ResNet-34
algorithm; the second step was to follow the eyes of the students and determine their
33
coordinates in the image using a trained model based on gaze tracking. The model was
trained on a (Gi4E) dataset that contains around 1380 images, split into 70% of the training
data and 30% of the test data . Through this, we could know the focus area of the student and
thus have the ability to warn the student if they are looking outside the computer screen, such
as looking at a cheat sheet, a book, or any other aid. The third step was determining the head
movement using a model with Haar cascade algorithms, which operate when the frontal face
is absent ; it was trained on a (FEI) face dataset that contains around 2800 faces belonging to
100 people, split into 70% of the training data and 30% of the test data. The fourth step was
detecting the presence of any material other than the face, for example, the presence of a
mobile phone, paper, book, or hand; the fifth step is detecting the presence of more than one
face in the image. The YOLO (V3) model, which was trained on the COCO dataset, was used
in the fourth and fifth steps to detect objects depicted in the camera (such as a human, book,
watch, iPad, and any other device) in real time , where this dataset contains 5000 images split
into 70% of the training data and 30% of the test data.

3.6. Data Analysis

Statistics were used to examine how various research findings can be handled and defined
mathematically, as well as how various inferences can be made from the outcomes of the
analysis. The datasets were examined using statistical analysis. We used the descriptive
analytic method to analyze the responses from these interviews. We organized the questions
and responses into themes. The results were represented as percentages

VI. Communicate appropriate project recommendations for technical and nontechnical

audiences

1. For non-technical audience:

This is an anti-cheating website in the academic field. Users just need to upload the text they
need to check, and the website will automatically search for parts that are plagiarized from
existing works. This will help you quickly find violations appearing in documents.

34
2. For technical audience:

With this website we have used APIs that are linked to huge data stores. through which you
can Initial Check or a Full Comparison.

35
36
VII. Discuss accuracy and reliability of the different research methods applied.

Step 1:

Based on expert suggestions and opinions during the development of the student
identification system, which verifies a student’s identity based on the algorithms used, it was
found that it is easy to compare the existing face in the database with the face that is taken by
the camera. In cases where the face is not recognized or matched, the results appear as
“unknown.” On the other hand, when the face is recognized, the name of the face appears as
it is saved in the database. Thus, the result of the performance and accuracy for the deep
metric learning algorithm and the ResNet-34 algorithm in this system is about 99.38%.

37
 Figure 5. Performance and accuracy of deep metric learning algorithms.

Adjustments were made after the experts provided their feedback about the perfor mance of
the system, as the experts and lecturers suggested that the system register the unknown
student in a state of absence directly within the system. The optimization was made, which
improved the efficiency of ResNet-34 and reduced the time required.

A system to identify students and track their bibliometric attendance was also de veloped by
to support educational platforms; however, the accuracy of that system was only 95.38%
compared to our system's 99.38%. In addition, our finding was closely related to the result
previously reported by , in which the authors show testing accuracy of 88.92% with a
performance time of fewer than 10 ms. This system was found to the suitable for identifying
individuals in CCTV footage in locations because it masked facial recognition in real-time.

Step 2:

38
The success of distance learning systems depends on many factors, the most important of
which is the extent to which lecturers understand the state of students' emotions and
behaviors during the educational session, in order for the lecturers to adjust the teaching
strategy in a timely manner to attract the attention of the largest number of students Another
factor is the ability of the lecturers to coordinate the students in classrooms during exams.
Thus, based on expert suggestions and opinions during the development of the system of
follow-up of the students in the class, it was developed to determine the feelings of the
students depending on the deep learning algorithms (CNN). This system takes many
consecutive photos for each student, shows the results graphically on screen, and also saves
them. Therefore, the system gives the lecturer the ability to review the results after every
session. It also allows the lecturer, in real time, to follow up with any student just by clicking
on the student’s name during the session.

When this system fails to find any faces, it notifies the student. This system is first in
identifying faces. After six notifications, either the student’s absence is noted for this session,
or the decision is ultimately forwarded to the lecturer. Second, the system tracks each
student’s facial expressions and emotions throughout the session. At the conclusion of each
session, the system saves the remarks and feedback (from interviews) for each student in the
form of a graph that displays the number of times they received messages, as well as other
information. Additionally, the lecture is free login capability whenever they want to view
student feedback. This system shows accuracy and high performance in discovering students’
emotions, where after training, the model achieved an accuracy of 66% on average (Figure
6), despite the way of expressing the same emotion differing from one student to another. The
authors in developed a system to discover students’ facial expressions and found that the
accuracy of the system did not exceed 51.28%. The system achieved 62% accuracy in student
follow-up during the online session. Our results outperformed previous results by 66%,
demonstrating improved student detection and verification in distance learning systems.
Generally, this helps lecturers in the online sessions to know students’ feelings effectively.

39
 Figure 6. System accuracy against time.

Moreover, changes made in response to the most recent expert advice suggest that the system
should provide post-lesson feedback to all students on the same graph in order to serve as a
session feedback chart according to the experts and lecturers. Experts and lecturers suggested
renaming some facial expressions or emotions with other names related to the student’s
mental state during the lesson, for example, changing the name of the emotional state of
sadness to “state of absent mind.”

Step 3:

The last section of this system addresses the most important factor in the success of distance
learning systems, which is monitoring students’ activities in e-exams. Thus, based on expert
suggestions and opinions during the development of the system, the system was building a
model to monitor students during the online exam by monitoring the iris of the eye based on a
gaze tracking algorithm, and the complete absence of the face, as it monitors the movement
of the entire eye. When the iris moves from the computer screen for more than five seconds,
the system will send an audio alert to the student. If the alert is repeated more than five times,
the student will be considered to be in a cheating situation. Similarly, if any of the following
cases are observed, such as head movement, the presence of any material other than the face,
or the presence of more than one face, the system will send an audio alert to the student.

40
Furthermore, the system makes a complete statistical profile for the situation of the student
during that time. For example, if no alert is sent to the student during the exam, the
framework will give a rundown after the end of the test that the student’s condition was
normal. Otherwise, when the student receives alerts, the system will provide feedback
showing the number of alerts and length of time of each abnormal case using the trained
model YOLO (V3).

Figure 7 displays the results of performance and accuracy in terms of students’ expressed
behaviors. This system demonstrated the accuracy and high performance in detecting the
abnormal behaviors of students in the e-exam, where the gaze tracking model achieved an
accuracy of up to 96.95%, and the facial movement tracking model achieved up to 96.24%.
In addition to that, the face detection and object detection models achieved an accuracy of
around 60%. Moreover, happy behavior achieved an accuracy of 45.27%, while fear achieved
an accuracy of 30.20%. From the lecturer’s point of view, this system helps teachers in online
exams to effectively identify expected cheating cases. Our findings were consistent with
results of earlier research that studied mechanisms that students use to cheat in online exams,
and their systems accurately (87.5%) tracked the movement of the student’s head, face, and
expressions of fear emotion during exams.

Figure 7. Performance and accuracy in terms of students’ expressed behaviors.

41
Adjustments were made following the experts’ feedback in which experts and lecturers
suggested sending an alert to the exam invigilators in real time for each abnormal behavior
(gaze, anger, facial movements, etc.) during e-exams, with a decision being made on the
returned alert by the exam invigilators. Furthermore, the experts and lecturers suggested
labelling each abnormal behavior descriptively in the feedback summary. At the end of the
feedback, the system was finally developed based on the opinions and needs of lecturers for
student invigilation. The opinions and comments of these reviewers (experts and lecturers)
summarized the importance of having such integrated systems in distance learning in
monitoring the students in e-exams. This is because the system was able to improve the
educational process and increase the credibility of the e-exams, as well as save the lecturer
time.

The quality of distance learning is ensured based on several factors, the most important of
which is the lecturer’s understanding of the needs of the educational tool or system that
supports them in delivering high quality invigilation. The system will efficiently recognize
students’ faces and emotions and verify and detect their attempts in e-exams. The system
developed here achieved this by improving the efficiency and credibility of e-testing systems.
It also helps detect cheating in exams by recording students’ feelings in real-time. This has
been suggested as a solution for the issues with distance learning in research.

VIII. Assess the extent to which the project recommendations meet the needs of the
identified organisation, including fullysupported rationales for planning decisions made
(M4).

The findings of this study revealed that the facial recognition system achieved high
performance and accuracy in detecting students’ expressed behaviors, abnormal behaviors,
gaze tracking, and facial movement tracking. The system was developed for invigilation
purposes in distance learning. The system was aimed to be used by university lecturers to
monitor students in e-exams, and it is equipped with a student verification system. Several
deep learning algorithms were applied to develop this system, and the objective was achieved
with high accuracy. However, during the development process, we faced challenges, such as
real time data for cheating detection. Hence, we relied on lecturers’ feedback in distance
learning systems. The review of previous studies also provided us with information for the
development of the system and analysis. Both forms of information were integrated to
improve the system’s efficiency and credibility for distance learning. In conclusion, this study

42
contributes knowledge of integrating different models to architecturally, imaginarily, and
statistically analyze image data using deep learning algorithms and an agile approach to
improve invigilation of students in distance learning systems. Although the effectiveness of
distance learning and the credibility of e-exams could be improved with the help of this
developed system, there were a number of drawbacks, such as the inability to automatically
detect head and face movements; the inability to account for the various ways that a single
student may express the same emotion to another; and the poor internet connection in rural
areas. Future research could include new capabilities to help students with unique needs and
speech detection to help the follow-up system during exams better identify cheaters.

IX. CONCLUSION:

In summary, undertaking a research project provides the opportunity to apply research and
analysis skills to draw meaningful conclusions. Through this process, I was able to
effectively apply qualitative and quantitative research. Evaluating the information collected
allows for informed decision making and identification of areas for improvement. Presenting
research findings effectively is important for communicating results and recommendations to
stakeholders. In addition, participating in teamwork will promote collaboration and exchange
of ideas, improving project results. Finally, reflecting on the results of the project allows for
the identification of lessons learned and opportunities for further development. By leveraging
these skills and approaches, research projects can yield valuable insights and contribute to
organizational growth and success.

43
X. REFERENCES

44