ScienceDirect

Available online at www.sciencedirect.com

ScienceDirect
Procedia Computer Science 00 (2022) 000–000
www.elsevier.com/locate/procedia
Available online at www.sciencedirect.com
Procedia Computer Science 00 (2022) 000–000
www.elsevier.com/locate/procedia
ScienceDirect
Procedia Computer Science 218 (2023) 1542–1550
International Conference on Machine Learning and Data Engineering

Detection and classification

International of anxiety
Conference on Machine inand
Learning university students
Data Engineering
through the application of machine learning
Detection and classification of anxiety in university students
a b*
through
Shaurya the application
Bhatnagar , Jyoti Agarwalof ,machine Sharmaa,b
learning
Ojasvi Rajeev
a
Dept of CSE, Amity University Uttar Pradesh, Sector 125, Noida, Uttar Pradesh 201313, India
a Deemed to be University,b* India a,b
Shaurya
b
Bhatnagar
Dept of CSE, Graphic Era, Jyoti Agarwal Dehradun, , Ojasvi Rajeev
Uttarakhand Sharma
248002,
a,b
a Dept of CSE, Amity UniversityUttar
UttarPradesh,
Pradesh,Sector
Sector125,
125,Noida,
Noida,Uttar
UttarPradesh
Pradesh201313,
201313,India
India
Dept of CSE, Amity University
b
Dept of CSE, Graphic Era Deemed to be University, Dehradun, Uttarakhand 248002, India
Abstract a,b
Dept of CSE, Amity University Uttar Pradesh, Sector 125, Noida, Uttar Pradesh 201313, India
Mental Health has recently transformed into a domain that caused interest in almost every field and has garnered attention in recent
Abstract
years, millions of people suffer due to mental illnesses that are often left unaccounted for, a vast majority of these people are
youngsters, a large chunk being students belonging to various universities. This research’s aim was to pinpoint the extent of anxiety,
Mental Health has
in conjunction, itsrecently
effects, transformed
as noticed in into a domain
Indian that caused
university interest
students. Theindataset
almost was
everygathered
field andusing
has garnered attention in
a questionnaire, thatrecent
was
years, millions of people suffer due to mental illnesses that are often left unaccounted for, a vast majority
matching with Likert scale measurement criteria, which consisted of university engineering students. This questionnaire was of these people are
youngsters, a large chunk
distributed amongst 127 being students
engineering belonging
students, as to variousthe
a result, universities. This research’s
level of anxiety aim was and
was quantified, to pinpoint the extent
its causes of anxiety,
and effects were
in conjunction,
identified, itsofeffects,
a series as noticed
statistical in and
reliability Indian university
validity students.
tests were The on
performed dataset was gathered
the dataset. Machineusing a questionnaire,
learning algorithms arethat was
applied
matching
in the end with Likertthe
to classify scale measurement
anxiety level basedcriteria, which consisted
on the effects of anxiety ofafter
university engineering
being trained students. data
on pre-existing Thispoints.
questionnaire was
It was found
distributed amongst alpha
that the Cronbach’s 127 engineering
value for thestudents, as a result,
entire dataset the level
was 0.723 of anxietycorrelation
and Pearson’s was quantified, and was
coefficient its causes
0.823,and
the effects
accuracy were
for
identified,
the a seriesdecision
naïve bayes, of statistical
tree, reliability and validity
random forest tests were
and support vectorperformed
machineon the dataset.
algorithms Machine
were learning
71.05%, algorithms
71.05%, 78.9% areandapplied
75.5%
in the end to classify the anxiety level based on the effects of anxiety after being trained on pre-existing data points. It was found
respectively.
that the Cronbach’s alpha value for the entire dataset was 0.723 and Pearson’s correlation coefficient was 0.823, the accuracy for
the2023
© naïveThe
bayes, decision
Authors. tree, random
Published forest andB.V.
by ELSEVIER support vector machine algorithms were 71.05%, 71.05%, 78.9% and 75.5%
respectively.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-
nd/4.0) Peer-review
© 2023 The Authors. under responsibility
Published by Elsevierof B.V.
the scientific committee of the International Conference on Machine Learning
©
and2023
This is anThe
Data openAuthors.
Engineering Published
access article underby theELSEVIER
CC BY-NC-NDB.V. license (https://creativecommons.org/licenses/by-nc-nd/4.0)
This is an open access article under the CC BY-NC-ND
Peer-review under responsibility of the scientific committee of the license (https://creativecommons.org/licenses/by-nc-
International Conference on Machine Learning and Data Engineering
nd/4.0) Peer-review under responsibility of the scientific committee of the International Conference on Machine Learning
and Data Engineering
Keywords: Anxiety; classification algorithms; dataset; machine learning; mental health; statistical analysis

Keywords: Anxiety; classification algorithms; dataset; machine learning; mental health; statistical analysis

1877-0509 © 2023 The Authors. Published by ELSEVIER B.V.

This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-
review under responsibility of the scientific committee of the International Conference on Machine Learning and Data Engineering
1877-0509 © 2023 The Authors. Published by ELSEVIER B.V.
This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) Peer-
review under responsibility of the scientific committee of the International Conference on Machine Learning and Data Engineering

1877-0509 © 2023 The Authors. Published by Elsevier B.V.

This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0)
Peer-review under responsibility of the scientific committee of the International Conference on Machine Learning and
Data Engineering
10.1016/j.procs.2023.01.132
 Shaurya Bhatnagar et al. / Procedia Computer Science 218 (2023) 1542–1550 1543
Author name / Procedia Computer Science 00 (2019) 000–000 2

1. Introduction

Anxiety happens to be a disorder which causes worry and nervousness so much that it begins to interfere with one’s daily
life. There can be several contributing factors that may trigger an individual and make them become anxious for prolonged
times, which can leave behind crippling and devastating after-effects, for e.g., feeling worthless, self-hatred, hopeless etc.
University students constitute a large portion of the sufferers, bearing numerous adverse effects caused by crippling anxiety.
There exist several questionnaires that are specifically tailored to conclude whether an individual suffers from anxiety and
its effects, such as the DASS-21 (Depression, Anxiety and Stress Scale), GAD-7 (Generalized Anxiety Disorder). The causes
and effects of anxiety vary from person to person, however certain circumstances can generalize the causes as well as the
effects of anxiety, this generalization can be attributed to several university students in India that is subject to copious
amounts of stress, this may be academic, social, or intrinsic stress. This paper has dived into the possibility of pinpointing
the plausible and common causes of anxiety, as well as linking said causes to certain effects that may surface in an average
Indian university student. A self-created question form was used to gather the dataset for this purpose, following which the
dataset gathered was subjected to certain statistical analyses to determine its reliability and validity, finally machine learning
classification algorithms were applied on the dataset to classify new data points into mild, moderate, or severe categories in
relevance to anxiety levels calculated using the identified causes. It has beennoticed that there has been a significant rise in
suicides in India especially during the past couple of years[1]. A study also found that 37 percent of University’s students in
India suffer from moderate depression [2]. This can be correlated with severe pressure that university students are often
labouring under. Several factors contribute towards the origin of this stress and are often disregarded assuming that the
student will simply get over it, however, this ordeal may render an individual crippled and unmotivated to extreme degrees.
The dataset collection included 127 engineering students that participated in this research, the questionnaire for this purpose
targeted numerous factors that is relatable to the causes and effects of anxiety that may present itself in an average Indian
university student. The students that participated in this research belong to an engineering background, however the
questionnaire designed can be used with students belonging to any study field at any respective University local to a certain
region. The Questionnaire consists of 40 questions that is divided into following categories:
• Demographic
• Probable Causes of Anxiety
• Probable Effects of Anxiety

1.1 Understanding Anxiety

Anxiety can be described as a feeling composed of worrisome thoughts, tension, and physical manifestations in the
form of high blood pressure and several other effects. There can be several reasons behind the feeling anxious, it is the
human body’s natural response to stressful situations, such as starting a new job, or one’s first day at college/university
[3]. It is essential to understand the difference between ordinary anxiety that manifests as a natural response, and
anxiety disorders that can adversely affect an individual’s life [4]. Ordinary anxiety is temporary, it is sporadic and
intermittent, usually kicking in when an individual is placed under stressful circumstances, for example, delivering a
presentation before an audience, or taking a huge test, the feeling of anxiety that accompanies these situations usually
subsides, and the individual returns to their original state of mind. However, anxiety disorders are perpetual. They
can severely incapacitate an individual, to an extent of rendering them incapable of carrying out everyday tasks. There
are several categories of anxiety disorders, as listed below [4]:
Panic Disorder: This causes an individual to experience a panic attack at any point in time, this may be triggered by
a situation and usually leaves the personally incapacitated.
Phobia: This is characterized by an intense fear of a certain situation or object, for example, claustrophobia is the fear
of being stuck in tight spaces, such as a small elevator.
• Social Anxiety Disorder: Characterized by the fear of social interactions, usually owing to the fear of
judgement by others.
• Obsessive Compulsive Disorder: Characterized by an individual compulsively and repetitively engaging in
a particular set of activities due to their intrinsic fears.
• Separation anxiety: Characterized by the fear of having to be away from loved ones.
1544 Shaurya Bhatnagar et al. / Procedia Computer Science 218 (2023) 1542–1550
Author name / Procedia Computer Science 00 (2019) 000–000 3

• Hypochondria: Being perpetually fearful of one’s health.

• PTSD (Post-traumatic stress disorder): Characterized by a feeling of anxiety that follows an extremely
traumatic event.

Anxiety manifests itself in the form of physical alterations in the body as well, a person experiencing about of
anxiety may exhibit the physical symptoms A feeling of dizziness, one may feel as though they are about to faint,
shortness of breath, dryness of mouth, increased blood pressure and heart palpitations, feeling nauseous etc[4]. Anxiety
disorders can affect a huge chunk of an individual's life, in 2017, 44.9 million people in India were suffering from
anxiety disorders [5].

1.2 Understanding Factors Affecting Anxiety

The feeling of worrying thoughts, mental pressure and accompanying physical signs and symptoms akin to cold
sweating, blood pressure, dry mouth and several other implications constitute anxiety. This feeling of anxiety can
present itself due to several reasons and situations, such as a job interview, or an exam [3]. Nevertheless, anxiety can
present a problem when it becomes perpetual and recurrent for an individual, therefore it becomes necessary to draw
out a distinction between ‘usual’ anxiety and disorders pertaining to anxiety that can be caused concern [4]. Anxiety
disorders can be devastating, considering that any individual suffering from its effects will find it extremely hard to
go about their daily routine without entering the vicious cycle of intrusive and worrying thoughts this may lead to
adverse and long and lasting effects on mental health.
The diagnosis of anxiety disorders does not narrow itself down to one single method, in this domain,
questionnaires present themselves as useful tools to judge the state of mind of an individual and their level of
anxiety, they consist of questions that directly highlight symptoms that an individual may be manifesting, this is
determining level of anxiety of an individual, GAD-7(General Anxiety Disorder), DASS-21 (Depression Anxiety And
Stress Scale ), PTSD questionnaire(Post-traumatic Stress Disorder), Social Anxiety Disorder questionnaire are some
of many questionnaires that are for this.
The objective of this paper was to quantify anxiety in University students and classify it into the mild,
moderate, andsevere category using machine learning algorithms by considering just the effects observed. The main
contribution this research aims to make is to target the causes and effects of anxiety in university studentsand classify
it to produce awareness about the potentially immense mental pressure faced by them.
This research paper has been organized into numerous This research paper has been organized into numerous
sections,section 2 elaborates on the work that has been done regarding past research related to the topic of this research,
section 3 describes the research methodology that has been used to achieve the objective of this paper. Section 4
consistsof the result analysis followed by conclusion and future scope in section 5.

2. Related Work

The objective of this paper was to review how different ML algorithms are used to diagnose different mental illnesses
and how they are employed. The algorithms reviewed are Support Vector Machine (SVM), Gradient Boosting Machine
(GBM), Naïve Bayes, Random Forest and K-Nearest Neighbors since they were in frequent use. It begins by defining
different types of machine learning algorithms into broad categories, i.e., supervised machine learning and unsupervised
machine learning, it then touches upon different algorithms being talked about throughout the research and their subtleties
and nuances. A research search engine is used to search for articles related to mental illnesses involving the use of machine
learning. The articles are then categorized in accordance with mental illness type, ML techniques, accuracy, and sample size.
Mental illnesses such as PTSD (Post Traumatic Stress Disorder), Schizophrenia, depression and autism are considered.
Different ML algorithms are applied in numerous research papers using various datasets that yield respective accuracies. It
was found out that most of the SVM classifiers yielded an accuracy of more than 75% in the papers that it was used [6].
Srividya, Mohanavalli and Bhalaji employ numerous machine learning algorithms, such as SVM (Support Vector Machine),
KNN (K-Nearest Neighbor), Logistic Regression, Naive Bayes Classifier to identify the state of mental health in a certain
target group. The questionnaire
 Shaurya Bhatnagar et al. / Procedia Computer Science 218 (2023) 1542–1550 1545
Author name / Procedia Computer Science 00 (2019) 000–000 4

consisted of a scoring system constituting a range 0-100, where 0-30 was labelled as mentally distressed, 31-60 as
neutral, and 61-100 as happy. All the algorithms performed almost equally well, and the use of ensemble machine
learning yielded an accuracy of 90% [7]. T. Wyatt, S. Oswalt and Y. Ochoa have performed the examination of
differences in the diagnosis related to mental health and the academic impact related to the same, considering the year
of college as a classification factor, they have done a secondary data analysis here where responses from US
Undergraduate students from the American College Health Association-National College Health Assessment II were
considered, the sample size was 66,159 students and regression analysis was performed. The results of this paper
outlined the distinction in diagnoses of mental health, found by classification, where it was reported that first year
students exhibited a higher rate of injury to self and serious consideration of suicide. The rest of the upper students a
heavier academic impact due to factors relating to mental health [8]. E. Garcia-Ceja, M. Riegler, T. Nordgreen, P.
Jakobsen, K. Oedegaard and J. Tørresen perform surveys related to recent research in the field of Mental Health
Monitoring Systems(MHMS), which uses Machine learning and sensor data, the focus has been placed on numerous
mental conditions such as anxiety, depression, stress, bipolar disorder etc, a classification structure has beenproposed
to direct the audit of the related work and the research challenges have been discussed. [9] Similarly, N. Mohd Shafiee
and S. Mutalib review mental health problems amongst higher education students in Malaysia, it talksabout the higher
education system in Malaysia and the factors affecting the students’ mental health, they have compared different
research papers in order to determine research gaps. The most common factors affecting the mental health were
identified as lack of social support, the learning environment and financial problems, it wasfound that related
research uses the SVM (Support Vector Machine) algorithm the most, and it yields an accuracy of 70% to 90% [10].
Another review performed by A. Le Glaz summarizes the studies in which machine learning and Natural Language
processing was utilized for the purpose of mental health, in addition to this, the potential use of these methods was
taken into consideration for clinical practice. 58 articles out a total of 327 were included in the review, it was observed
that population studies could be widely grouped into 3 categories: social media users, patients included in medical
databases and patients coming into emergency. Python was used most frequently as the language [11]. P. Tiffin and
L. Paton explore the benefits of applying machine learning to mental healthcare in addition to conventional statistical
modelling, the challenges and limitations have been discussed as well. They have described machine learning as a tool
in the box of several others that can pave the way to real time practice [12].
M. Dobias, M. Sugarman, M. Mullarkey and J. Schleider find out if adolescents battling depression had access to mental
health treatments. They used symptoms of depression and sociodemographic variables for this purpose. They performed a
secondary data analysis; the dataset was from the National Survey of Drug Use and Health. 4 cross-validated random forest
models were used to predict whether depressed adolescents with symptoms that were elevated had accessed treatment in the
prior 12 months, it was found that 53.38% of them had accessed some form of treatment, it was found that neither the
symptoms of depression, nor the socioeconomic variables could predict the access of treatment, they also go on to highlight
the limitations of the research [13]. Y. Liu, J. Hankey, B. Cao and P. Chokka screen for MDD (Major Depressive Disorder)
in tertiary care environments. They use machine learning to test their composite screening application, which involves early
detection and life history factors, the primary aim in this study was to improvespecificity in a tertiary setting. There was a
total of 955 participants in this study that went through an interview with a psychiatrist, here, machine learning was used to
make more accurate predictions on an individual level, the accuracy of their tool was 72%, the sensitivity was 74.2%, and
the specificity was 69.8% [14]. A. Sheik Abdullah, R. Abirami, A. Gitwina and C. Varthana examine the psychotherapeutic
support offered to students to overcome distress during the COVID-19 pandemic, it does so by using machine learning
techniques. It evaluates the effectiveness of the academic performance of the students during the situation of the pandemic
for this purpose. They have taken a hybrid approach using a generic algorithm and upon statistical evaluation, they have
used an artificial neural network. The three key factors taken into consideration are cognitivism, behaviorism and social
constructivism. The dataset included 775 students. The data was processed by the algorithm and then classified using the
neural network. The hybrid approach yielded an accuracy of 88.18%, precision 94.69%, recall 92.94%. The P value for the
Fisher’s F-Statistical test was P < 0.001 [15]. F.M. Albagmi, A. Alansari, D.S.A Shawan, H.Y. AlNujaidi and S.O. Olatunji
leverage the power of machine learning by classifying two-class and three-class anxiety issues at an early stage during the
covid-19 pandemic, the dataset was collected via an online survey and consisted of 3017 participants, they identified the
factors affecting anxiety levels, followed by responses from the GAD-7 questionnaire, following which they used the
Support Vector Machine classifier which yielded promising results [16]. K.-S. Na, S.-E. Cho, and S.-J. Cho follow an
approach based on machine learning, they aim to draw out a distinction between panic disorder and other anxiety disorders
by taking
1546 Shaurya Bhatnagar et al. / Procedia Computer Science 218 (2023) 1542–1550

HRV(Heart Rate Variability) into consideration, the dataset consisted of 60 patients diagnosed with panic disorder and 61
others diagnosed with other anxiety disorders, it was seen that logistic regression performed best with an accuracy of 78.4%
[17]. W. A. Bokma et al., focus on the recovery period of anxiety, and predict it using machine learning algorithms, they
used a dataset consisting of 887 patients with different forms of anxiety disorders, several baseline predictors were used to
predict the recovery process, the random forest algorithm was used in this case and it was found that 484 patients had
recovered from their anxiety disorders [18]. V. Khullar, R. G. Tiwari, A. K. Agarwal, and S. Dutta focus on
GAD(Generalized Anxiety Disorder) and enumerate an automatic intelligent system to discern anxiety based off of
physiological symptoms. The dataset was collected and pre-processed a comparison was drawn out between ML
algorithms and ensemble learning [19]. I. C. Passos et al., zero in on the probability of suicide attempts amongst
patients with mood disorders, the dataset included a total of 144 patients that were diagnosed with mood disorders,
the machine learning model was trained on the basis of demographic variables, the subjects were then categorized into
suicide attempters or non-attempters, the accuracies lied between 65% to 72% [20]. L. Jani Anbarasi, M. Jawahar, V.
Ravi, S. M. Cherian, S. Shreenidhi & H. Sharen aim to analyze sleeping disorders and anxiety disorders amongst a
wide variety of individuals belonging to different strata of life in view of the recent COVID-19 pandemic, the study
also aims to predict symptoms in association with stress induced due to the pandemic. A total of 740 participants were
included in the research, GAD-7 and PSQI index were used to measure stress. The study highlighted the connection
between a lower social status and mental health problems as well. [21]

3. Experimental Setup

The following software was used for the purpose of this study:

1. IBM SPSS: A software developed by IBM for the purpose of data analysis.
2. Jupyter Notebook: A server-client application that helps the user to run and edit notebook documents through a
web browser.
In addition to the above, the following python libraries were used:
3. NumPy: A powerful python library that is used for the purpose of working with multi-dimensional arrays and
matrices.
4. Matplotlib: A cross-platform library that is used for the purpose of data visualization and the plotting of graphs,
usually used in conjunction with NumPy.
5. Pandas: A software library used for the purpose of manipulation and analysis of data in a tabular form.
4. Scikit-learn: A robust software library that offers a wide variety of tools for the purpose of implementing
machine learning algorithms such as regression and classification algorithms.

4. Research Methodology

From the related work, it can be observed that numerous statistical analysis techniques as well as machine learning
algorithms have been used to deduce meaningful results in terms detection, prediction and diagnosis of mental health
issues. The complete procedure that was followed for the purpose of this study is mentioned in Fig. 1.

Step 1: Identify factors that affect the mental health and well-being of Indian engineering students: In order to
achieve this task, an in-depth literature review was done, and numerous factors that affect the mental health of
university students were deduced and taken into consideration, each individual research work was scanned, and
various pre-existing questionnaires such as the GAD-7(General Anxiety Disorder), DASS-21 (Depression Anxiety
And Stress Scale ) were referred to as well. In addition to this, a psychologist was also consulted for the purpose of
the identification of the factors.
 Shaurya Bhatnagar et al. / Procedia Computer Science 218 (2023) 1542–1550 1547

Fig. 1. Research Workflow

Fig. 1. Research Workflow
Fig. 1. Research Workflow
Step 2: Develop an original questionnaire Fig.the
for 1. Research
purpose Workflow
of data collection: After taking various factors that
Step Fig.the
1. Research Workflow
affect2:
Step theDevelop
2: mental health
Develop
an original
an original
questionnaire
of engineering
questionnaire
for
university for
purpose
students
the
of data
into consideration,
purpose of
collection:
an originalAfterquestionnaire
taking various wasfactors
developed that
Step
affect 2:
theDevelop
mental an original
health of questionnaire
engineering university for students
the purposeinto of data collection:
data
consideration, collection:
an
After taking
After
original
taking various
questionnaire
various
was
factors
factors
developed
that
that
which2:
Step
affect has
the 40 questions
Develop
mental health targeting
an original
of the potential
questionnaire
engineering university causes
for the as well
purpose
students into asconsideration,
effects
of dataof anxiety
collection:
an experienced
Afterquestionnaire
original taking byvarious
Indianwas engineering
factors
developed that
affect
which the
hasitmental
40 health of
questions engineering
targeting university students intoasconsideration, an original questionnaire wasengineering
developed
students,
affect
which the
has carefully
mental
40 considers
health
questions of allthe
engineering
targeting the
potential
aspects university
potential
causes
surrounding asuniversity
students
causes as
well effects
life and
intoasconsideration,
well effects
of narrows
of
anxiety experienced
down questionnaire
an original
anxiety potential
experienced
by sources
by
Indian
Indian of stress to
wasengineering
developed
which hasit40
students, questions
carefully targeting
considers the potential causes asuniversity
well as effects of narrows
anxiety experienced by sources
Indian engineering
a few common
which
students, hasit40 parameters
questions
carefully thatall
targeting
considers all
aspects
affect
the
surrounding
a large
potential
aspects chunk
causesofasuniversity
surrounding students, life and
such
well as effects
life as,ofsocial
and anxiety
narrows life,down potential
academics,
experienced
down potential byfamily
Indianlife,
sources
offinancial
stress to
engineering
of stress
students,
asituation
few common it carefully considers
parameters that all aspects
affect a surrounding
large chunk of university
students, life
such and
as, narrows
social life,down potential
academics, sources
family life, stress to
offinancial to
students,
aa few etc.
common This
it carefully questionnaire
considers
parameters that was
all
affect distributed
aspectsa surrounding
large chunkamongstof studentssuch
university
students, andand
life garnered
as, narrows
social 127
life, responses.
down potential
academics, The questions
sources
family on the
offinancial
life, stress to
few common
situation etc. This parameters that affect
questionnaire was a large chunk
distributed amongstof students,
students such
and as, social 127
garnered life, responses.
academics,The family life, financial
questions on the
survey
asituation follow
few commonetc. a questionnaire
cause and
parameters that effect
affect arelationship,
large chunk which
of consists
students, such of
as, questions
social life, targeting
academics,The the potential
family life,causes
financial of
survey etc. This
situationfollow This a “Iquestionnaire
cause
was
was distributed
and effect distributed
relationship,
amongst
amongstwhich
students
students
consists
and
and of garnered
garnered
questions
127
127 responses.
responses.
targeting The questions
questions
thepotential
on
on the
causes the
ofa
anxiety,
situation
survey such as,
etc. This get distressed/anxious
aa questionnaire due
was distributed amongst to my friends and/or
students their
and of behaviour.”,
garnered 127 responses. this questions
The suggests
questions that
on the
survey follow
anxiety, follow
such as, “I
cause
cause
get
and
and effect
effect relationship,
distressed/anxious relationship,
due to my
which
which
friends
consists
consists
and/or of questions
their questions
behaviour.”,
targeting
targeting
this
the
thepotential
questions potential
suggests
causes
causesthat
of
ofa
student
survey
anxiety, may
follow
such beas, anxious
a “Icause
get due
andto effect
the company
distressed/anxious they
relationship,
due to arewhich
my keeping,
friends similarly
consists
and/or there
of questions
their are questions
behaviour.”, targeting
this that
thetarget
questions potentialthe causes
suggests potential of
anxiety,may
student suchbe as,anxious
“I get distressed/anxious
due to the company duethey to myare friends
keeping, and/or theirthere
similarly behaviour.”,
are this questions
questions that targetsuggests
the that aa
that
potential
effects of
anxiety,
student thisbe
such
may anxiety
as, “I getthat
anxious duemay manifest
distressed/anxious
to the
the company induea they
student,
to my such as,and/or
are friends
keeping, “I can’ttheir
similarly seem toareenjoy
behaviour.”,
there things which
this questions
questions that I suggests
target used
the to before
that a
potential
student
effects may beanxiety
anxious due to company in a they are keeping, similarly there are questions that target the potential
such as of
student
effects TV,
may
of
this
thisgames,
beanxiety
that
movies,
anxious
may
duemay
that
manifest
music,
to the etc”, these
company
manifest in a
student,
questions
they
student,
such
helpas,
are keeping,
such us“I
as,
can’t seem
understand
similarly
“I can’t there
seem theto
toare
enjoy
impact things
that this
questions
enjoy things
which
that anxiety
which
I used
target
I used
to
isthemaking.before
potential
to before
effects
such as of TV,this anxiety
games, that may
movies, manifest
music, in a student,
etc”, these questions such
helpas,us“Iunderstand
can’t seemthetoimpact enjoy things
that this which
anxietyI used to before
is making.
effects
such as of this anxiety that may manifest these in a student, such as, “Iunderstand
can’t seem toimpact enjoy things which I used to before
such as TV,
TV, games,
games, movies,
movies, music,
music, etc”,
etc”, these questions
questions help help us
us understand the the impact that that this
this anxiety
anxiety is is making.
making.
Step as
such 3: TV,
Datagames,
pre-processing:
movies, music, This etc”,
is onetheseof the most crucial
questions help ussteps to follow,thetheimpact
understand data that
thatwasthisgathered
anxiety isconsisted
making. of
Step
a small 3: number
Data pre-processing:
of missing values Thisthatis were
one ofnot theextremely
most crucial steps totofollow,
significant the datatherefore
the research, that wasthose gathereddata consisted
points were of
Step 3: Data
Data pre-processing: This is one of the most crucial steps to follow, the data that
that was gathered consisted of
aStep
not small3: number
considered. pre-processing:
of addition
In missing to valuesThis
this, that
datais were
one ofnot
theextremely
transformation mosthas crucial steps
significant
been done tothat
tofollow,
the the the
research,
converts datatherefore
string was
data gathered
those data consisted
responses points
to were
numeric of
Step
aa small 3: number
Data pre-processing:
of missing valuesThisthatis were
one ofnot theextremely
most crucial steps totofollow,
significant the the datatherefore
research, that wasthose gathereddata consisted
points wereof
notsmall
data number
considered.
adhering of
In missing
addition values
to this, that
data were not
transformation extremely has significant
been done to
that the research,
converts the therefore
string data those data
responses points
to were
numeric
anot
small numbertoIn
considered. ofthe Likert to
missing
addition Scale,
values
this, for
that
data example,
were not“Strongly
transformation extremely has Agree”
significant
been donewasthatassigned
to the a numeric
research,
converts the value
therefore
string data of
those “5”, andpoints
data
responses to“Strongly
were
numeric
not
data considered.
adhering
Disagree” wasto Intheaddition
assignedLikert to this, for
Scale,
a numeric data transformation
example,
value of “1”.“Strongly hasAgree”
In addition been done
to this, that
wasfor converts
assigned
some the string
a numeric
questions, dataof
anvalue
inversionresponses
“5”,ofand to“Strongly
scales numeric
had to
not
data considered.
adhering to Intheaddition
Likert to this, for
Scale, data transformation
example, “Strongly hasAgree”
been donewas that converts
assigned the string
aa numeric dataof
value responses
“5”, and to“Strongly
numeric
data adhering
Disagree”
be implemented was to the
assignedLikert
for Likert
them a Scale,
numeric
to adherefor example,
value
to of
the “1”. “Strongly
pattern In additionAgree”
of anxiety to was
this,
level forassigned
some
calculation. numeric
questions,
The datasetanvalue of
inversion
was “5”,
thenofand
scales
split “Strongly
hadthe
into to
data adhering
Disagree” was to the
assigned a Scale,
numeric for example,
value of “1”. “Strongly
In additionAgree”
to was
this, forassigned
some a numeric
questions, anvalue of
inversion “5”, ofand
scales“Strongly
had toto
Disagree”
be implemented
training was
setwas assigned
andassignedfor
test them
set, a numeric
to
where adhere value
70%value to
oftothe of
the “1”.
pattern In addition
of anxietyto this,
level for some
calculation. questions,
The an
dataset inversion
was thenof scales
split had
into the
Disagree”
be implemented for them a numeric
to adhere ofdata
the “1”.points
pattern of were
In addition considered
anxietyto this,
level for for
some thequestions,
calculation. training
The setanand
dataset 30%then
inversion
was ofofthe datainto
scales
split points
had to
the
be implemented
training
were set and test
considered forset,
for them
the toset.
where
test adhere
70% oftothethedata
Standardization pattern wasof
points anxiety
were level
for thecalculation.
considered of The
for the training datasetandwas
setscaling. 30%then of thesplit
datainto the
points
be implemented
training set them to adhere to thedata pattern ofalso done
anxiety level process
calculation. feature
The dataset was then split into the
training set and
were considered and testtest set,
for set,
where
where
the test
70%
set.70%
of
of the
the data points
Standardization points
was also
were considered
weredone
considered for
for the
for the process the training
training
of feature
set and
and 30%
setscaling. 30% of of the
the data
data points
points
training
were set and test
considered for set,
the where
test set.70% of the data points
Standardization was weredone
also considered
for the for the training
process of feature setscaling.
and 30% of the data points
were considered for the test set. Standardization was also done for the process of feature scaling.
Step 4:
were Reliability
considered for Analysis:
the test set.The Cronbach’s alpha
Standardization was alsotest was
doneperformed
for the process for the ofpurpose of reliability analysis of the
feature scaling.
Step 4: Reliability
questionnaire. This was Analysis:
done toThe checkCronbach’s
whether the alpha test was performed
questionnaire for the purpose
was consistently of reliability
able to measure analysis of the
the unidimensional
Step
Step 4:
4: Reliability
Reliability Analysis:
Analysis: The
The Cronbach’s
Cronbach’s alpha
alpha test
test was
was performed
performed for
for the
the purpose
purpose of
of reliability
reliability analysis
analysis of
of the
the
questionnaire.
factor This was done to check whether the questionnaire was consistently able to measure the unidimensional
Step 4:ofReliability
anxiety
questionnaire. This levels.
was The
Analysis:
done Cronbach’s
The
to checkCronbach’salpha test
whether alpha
the happens
test wasto be
questionnaire one
performed
was of thefortests
consistently the which
purpose
able is used
to to determine
of reliability
measure the analysisinter-item
unidimensional of the
questionnaire.
factor of
reliability anxiety This
and consistency, was
levels. done
The to check
Cronbach’s
especially whether
alpha the
test questionnaire
happens to be was
one consistently
of the tests able
which to
is measure
used to the unidimensional
determine inter-item
questionnaire.
factor of This was The
done to checkgiven whether that thehappens
the survey been
questionnaire put
wasby mases
consistently ofwhich
the analysis
able to and
measure subjected to
the unidimensional test has
factor
reliability of anxiety
anxiety
been tailored
levels.
levels.
andtoconsistency,
Likert The Cronbach’s
Cronbach’s
especially given
alpha
alphathat
test
testthehappens
survey
to
to be one
be put
been one by
of
of the
the tests
mases testsofwhich
thethe
is
is used
analysisused to
to determine
and determine
subjected
inter-item
inter-item
to test has
factor of anxiety
reliability and levels.Scale
consistency, measurement.
Theespecially
Cronbach’s givenalphaThere
thattestthe issurvey
a mathematics
happens been oneinvolved
to be put byof mases
the tests with
of which
the explanation
is usedand
analysis of the Cronbach’s
to determine
subjected inter-item
to test has
reliability
been
alphatailored and
tailored
test, and toconsistency,
Likert
it istoconsistency, Scale
characterized especially
measurement.
by given
the formula that
There
given the issurvey
a been
mathematics put by mases
involved of
with the analysis
the and
explanation subjected
of the to test
Cronbach’s has
reliability
been Likert Scale especially
measurement. given thatThere thebelow:
issurvey
a been put involved
mathematics by mases with of thethe analysis and subjected
explanation of the to test has
Cronbach’s
been
alphatailored
test, it istocharacterized
Likert Scale by measurement.
the formula There given below:is a mathematics involved with the explanation of the Cronbach’s
been
alphatailored istocharacterized
alpha test, Likert Scale by measurement. There is a mathematics involved with the explanation of the Cronbach’s
test, it
it is characterized by the
the formula
formulaሻȀሺ˜+
ƒൌሺ∗
given
given below:below:
((k − 1)∗ ሻሻ
alpha test, it is characterized by the formula given below: (1)
ƒൌሺ∗ ሻȀሺ˜+ ((k − 1)∗ ሻሻ (1)
ƒൌሺ∗ ሻȀሺ˜+ ((k
ƒൌሺ∗ ሻȀሺ˜+ ((k − 1)∗ ሻሻ − 1)∗ ሻሻ (1)
ƒൌሺ∗ ሻȀሺ˜+ ((k − 1)∗ ሻሻ (1)
where c is average of all covariances between items, k is the number of scale items and v denotes (1) the average variance of
where c is A
each item. average
score for of all covariances
every one itembetween
of the scale items, k is the number
is calculated, and a of scale itemsis and
comparison drawv denotes
with thethe average variance of
where c is average of all covariances between items, k is the number of scale items and v out
denotes the total score
average for each
variance of
whereitem.
each c is A
average
score of all
for covariances
every one item between
of the items,
scale is k is the number
calculated, and a of scale itemsis and
comparison draw v out
denotes
with the
the average
total variance
score for eachof
observation
where
each c is A
item. that
average
score theforscale
of all consists
covariances
every of inbetween
one item oftothe
calculate
items,
scale isthek Cronbach’s andalpha,
is the number
calculated, a of this is
scale
comparison then
items compared
is and
draw v out
denotesto the
with thevariance
the average
total scorevariance
for eachof
each item. Athat
observation score the for every
scale one item
consists of inoftothe scale isthe
calculate calculated,
Cronbach’s andalpha,
a comparison
this is thenis draw
comparedout with
to thethevariance
total score for each
& theitem.
each individual
observation Athat
score theitem
for
scalescores.
every oneTo
consists of interpret
item inoftothe the is
scale
calculate Cronbach’s
calculated,
the Cronbach’s alpha
andalpha,analysis
this is result,
a comparison thenis drawtheout
compared purpose
with
to the theof some
total
variance parameters
score for each
& observation
the to that theitem
individual scalescores.
consistsToof interpret
in to calculate
thefalls the Cronbach’s
Cronbach’s alpha, this is result,
then compared to the of variance
& needs
observation
the be that
understood.
individual theitem The
scalescores. alpha
consists To coefficient
of interpret
in to calculate
the thewithin thealpha
Cronbach’s
Cronbach’s range
alpha
analysis
alpha, ofthis
0 tois1,result,
analysis 0 shows
then
the that
compared
the
purpose
thethe
to
purpose
some relatability
inter-item
variance
of some
parameters
parameters
& the individual item scores. To interpret the Cronbach’s alpha analysis
needs to be understood. The alpha coefficient falls within the range of 0 to 1, 0 shows that the inter-item result, the purpose of some relatability
parameters
& the to
needs individual
be understood.item scores.
The To interpret
alpha coefficient thefallsCronbach’s
within thealpha
range analysis
of 0 to result,
1, 0 showsthe that
purpose
the of some relatability
inter-item parameters
needs to be understood. The alpha coefficient falls within the range of 0 to 1, 0 shows that the inter-item relatability
needs to be understood. The alpha coefficient falls within the range of 0 to 1, 0 shows that the inter-item relatability
1548 Shaurya Bhatnagar et al. / Procedia Computer Science 218 (2023) 1542–1550

is absent and that the items on that scale are completely independent of one another, which also signifies that there is
no reliability. On the hand, if the alpha value deduced were to approach 1, then that would mean that the inter -item
relatability is high, which shows that the consistency of the results is high, and that the scale is highly reliable [20].
Alpha approaches one if the number of items on the scale approach infinite. Another way of looking at it is, if the
extent to which the items on the questionnaire are measuring the same variable needs to be known, in which case, ‘1’
would mean that all the items on the scale consistently measure the same thing and ‘0’ would mean all the items are
independent and measure completely different things altogether. However, 0 and 1 are ideal values, and the real- world
values of alpha lie between 0 and 1, a “good” alpha value completely depends on the scale being used and what it
aims to measure however, it is said that a good alpha value should lie within the range of 0.65 to 0.8(or higher). If the
scale that is being analyzed is unidimensional then any values below 0.5 should be immediately rejected. It should be
noted that the alpha value is a function of both the number of items on the scale and the covariances, this means that
the alpha value can be increased by simply increasing the number of items present on the scale, therefore it should be
noted that a very high alpha value is subject to redundancy in the items. For this research, IBM SPSS software was
used in order to perform both the reliability and validity analysis.

Step 5: Validity Analysis: This analysis was performed between 2 variables, the anxiety calculated using the cause-
based questions and the overall effects of anxiety. In general, a link between two items can be identified as the
correlation between them, the value denoting that correlation is said to be within the range of -1 to +1, this is also
known to be the Pearson’s correlation coefficient (r) when measured between two continuous variables, a direct
relationship is characterized by the positive value of the coefficient and an inverse relationship of any two variables
is characterized by a negative value, however, the value 0 is indicative of the fact that there is absolutely no linear
relationship that is there between the two variables being subjected to the test, although useful, this should not be
construed in a sense that other forms of non-linear relationships do not exist between the two variables.

Step 6: Applying various classification algorithms for the purpose of anxiety classification: The dataset was
separated in 3 classes of anxiety levels, i.e., mild, moderate, and severe. This was done according to the following
mean range of anxiety calculated using the cause-based questions:

• Mild: 1 to 2.3
• Moderate: 2.3 to 3.6
• Severe: 3.6 to 5

After this, the new data-points were attempted to be classed given the values of effects-based questions, that is, the
severity of the effects of anxiety was used as training data to correctly class the new data-points. The questions that
target the effects of anxiety were split into the training and test set in a 70% and 30% ratio. The algorithms applied
were the Naïve Bayes, Random Forest, Decision Trees, and SVM (Support Vector Machine).

5. Result Analysis

The reliability and validity analyses followed by the application of Machine Learning algorithms which were applied
to dataset yielded certain significant results which are discussed in this section. While performing a reliability analysis
on the whole dataset, an alpha value of 0.723 was achieved. Furthermore, a reliability test on the questions that pinpoint
the causes of anxiety yielded an alpha value of 0.871, similarly, questions that pinpoint the effects of anxiety yielded
an alpha value of 0.843. This is shown in Table 1. This provided preliminary evidence of the following fact: The
questions on the questionnaire reliably and consistently measures anxiety levels, since our Cronbach's alpha value
falls within the range of 0.7 to 0.9. The Cronbach’s alpha test consistently yielded a value above 0.7, which is indicative
of the fact that this questionnaire consistently measures the underlying factor of anxiety calculated using the cause-
based questions, in addition to this, a bivariate correlation (validity analysis) is performed between levels of anxiety
deduced from the cause variables and overall effects of anxiety (average). The validity analysis yielded a Pearson
correlation value of 0.823. SIG 2 tailed value for the validity analysis is 0.001 (< 0.05) which indicates a significant
directly proportional relationship. The validity test, namely, the Pearson’s correlation analysis proved that the
relationship between average anxiety pertaining to all data-points and overall effects of anxiety holds a positive,
directly proportional relationship, that is, if the causes rise, the effects of anxiety will also rise. The application of ML
algorithms can be interpreted as observed by the accuracy levels. The accuracies obtained are shown in Table 2. Fig.
2 depicts the accuracies yielded by different machine learning algorithms applied. The naïve bayes classification
algorithm yielded an accuracy of 71.05 %, that means 71.05 % of
 Shaurya Bhatnagar et al. / Procedia Computer Science 218 (2023) 1542–1550 1549
the data-points were correctly classed after being trained on the data provided, which consisted of the effects of anxiety
the data-points were correctly classed after being trained on the data provided, which consisted of the effects of anxiety
questions and the level of anxiety calculated using the questions on causes, and the random forest algorithm, decision
questions
the and the
data-points level
were of anxiety
correctly calculated
classed using
after being the questions
trained on provided,
on the data causes, and the random
which forest
consisted algorithm,
of the effects ofdecision
anxiety
tree & SVM algorithms yielded accuracies 78.9 %, 71.05%, 75.5%. The random forest algorithm appears to have
tree & SVM
questions andalgorithms
the level ofyielded
anxietyaccuracies
calculated 78.9
using%,the71.05%,
questions75.5%. The and
on causes, random forest algorithm
the random appears decision
forest algorithm, to have
performed the best. The computation time for the SVM, Random Forest, Naïve Bayes and Decision trees algorithm
performed
tree & SVM thealgorithms
best. The computation time for
yielded accuracies the%,
78.9 SVM, Random
71.05%, Forest,
75.5%. TheNaïve
randomBayes andalgorithm
forest Decisionappears
trees algorithm
to have
was noted as 0.016s, 0.048s, 0.028s and 0.013s respectively.
was noted as
performed the0.016s, 0.048s,
best. The 0.028s and
computation 0.013s
time respectively.
for the SVM, Random Forest, Naïve Bayes and Decision trees algorithm
was noted as 0.016s, 0.048s, 0.028s and 0.013s respectively.
Table 1. Alpha values for different variables.
Table 1. Alpha values for different variables.
TableVariables
Variables
Considered
1. Alpha values Alpha Value
for different variables.
Considered Alpha Value
All Variables
Variables Considered 0.723 Value
Alpha
All Variables 0.723
Cause-based variables
All Variablesvariables 0.871
0.723
Cause-based 0.871
Effect-based variables
Cause-based variables 0.843
0.871
Effect-based variables 0.843
Effect-based variables 0.843
Table 2. Accuracies for classification algorithms.
Table 2. Accuracies for classification algorithms.
Algorithm
Table Accuracy
2. Accuracies for classification algorithms.
Algorithm Accuracy
Decision Trees
AlgorithmTrees 71.05%
Accuracy
Decision 71.05%
Naïve Bayes
Decision Trees 71.05%
71.05%
Naïve Bayes 71.05%
Random Forest
Naïve Bayes 78.9%
71.05%
Random Forest 78.9%
Support
Random Vector
Forest Machine 75.55%
78.9%
Support Vector Machine 75.55%
Support Vector Machine 75.55%

Fig. 2. Graph Representing Accuracy

Fig. 2. Graph Representing Accuracy
Fig. 2. Graph Representing Accuracy
6. Conclusion and Future Scope
6. Conclusion and Future Scope
6. Conclusion and Future Scope
The result of this study yields a wide variety of meaningful conclusions. The reliability and validity analyses provide
The result of this study yields a wide variety of meaningful conclusions. The reliability and validity analyses provide
aThestrong foundation
result for the
of this study correctness
yields of the questionnaire
a wide variety of meaningfuland the factorsThe
conclusions. they aim to analyze. The reliability analysis
a strong foundation for the correctness of the questionnaire and the factors theyreliability and validity
aim to analyze. analyses analysis
The reliability provide
is indicative
aisstrong of
foundationconsistency in the
for the correctness measurement of anxiety
of the questionnaire levels,
andlevels, that
the factors is,
theyit consistently measures those levels of
indicative of consistency in the measurement of anxiety that is, it aim to analyze.
consistently The reliability
measures analysis
those levels of
anxiety
is beingof
indicative calculated. In addition
consistency to this, the validity
in the measurement analysis
of anxiety indicates
levels, that aais,positive relationship
it consistently between
measures the measured
those levels of
anxiety being calculated. In addition to this, the validity analysis indicates positive relationship between the measured
anxiety levelscalculated.
anxiety being and overall effects of to
anxiety, thevalidity
evidence here is clear: with the rise of causes of anxiety,thethe overall
levels and overallIneffects
addition this, the
of anxiety, the evidenceanalysis
here isindicates a positive
clear: with relationship
the rise of causes of between
anxiety, measured
the overall
effects
anxiety will also
levels be
and aggravated, this is a powerful conclusion that can be drawn and used as a strong pillar for further
effects will also beoverall effects
aggravated, of isanxiety,
this the evidence
a powerful conclusion here is clear:
that can bewith
drawn theand
riseused
of causes of anxiety,
as a strong pillar the
for overall
further
research,
effects causes
will of
also be anxiety can
aggravated, be pinpointed
this and
is a powerful linked to
conclusion the effects, which will eventually pave a path to mitigation.
research, causes of anxiety can be pinpointed and linked to thethat can be
effects, whichdrawn willand used as apave
eventually strong pillar
a path for further
to mitigation.
Apart from
research, this, of
causes the machine canlearning algorithms
and performed almost equally
whichwell, will averaging
eventuallyaapave
75%aaccuracy, however
Apart from this, theanxiety
machine be pinpointed
learning algorithms linked to the
performed effects,
almost equally well, averaging path to mitigation.
75% accuracy, however
the random
Apart forest algorithm performed best, yielding an accuracy of 78.9%, which alludes to the fact that the severity
the random forest algorithm performed best, yielding an accuracy of 78.9%, which alludes to the fact that thehowever
from this, the machine learning algorithms performed almost equally well, averaging a 75% accuracy, severity
oftheeffects
randomof anxiety can be performed
a significant factor that can be used of to classifywhichanxiety into mild,
to themoderate andseverity
severe
of effects offorest algorithm
anxiety can be a significant best, yielding
factor an accuracy
that can be used to 78.9%,
classify anxiety alludes
into mild, fact that the
moderate and severe
categories,
of this
effects ofthis information
anxiety wields
can bewields a lot
a significant of power, in that it can help pinpoint several students that may be at risk of a
categories, information a lot offactor
power,that can itbecan
in that used to pinpoint
help classify anxiety into mild,that
several students moderate
may be and
at risksevere
of a
severe mental
categories, breakdown.
thisbreakdown.
informationIn In the
wields future, this
a lot ofthis research
power, can
in that be
it can extended to
help pinpoint various other
severalother departments
students apart
that mayapart from
be atfrom thea
risk of
severe mental the future, research can be extended to various departments the
engineering
severe mental and technology department
engineering andbreakdown.
technologyIndepartment
the future, this research can be extended to various other departments apart from the
engineering and technology department
1550 Shaurya Bhatnagar et al. / Procedia Computer Science 218 (2023) 1542–1550

to help understand the causes and effects of anxiety amongst students who are studying different courses, moreover,
the research can be extended to other locations in the country as well to understand what impact different regions have
and what challenges students living in these regions to face, as well as how these challenges and anxiety triggers differ
from one another, so as to cater to each subset of students in the most competent and effective manner.

References

[1] Swain, Prafulla Kumar, Manas Ranjan Tripathy, Subhadra Priyadarshini, and Subhendu K. Acharya (2021) “Forecasting suicide rates in
India: An empirical exposition”. PLOS ONE 16(7):1-21.
[2] Sawhney, Manisha, Seth Kunen, and Ashum Gupta (2018) “Depressive Symptoms and Coping Strategies Among Indian University Students.”
Psychological Reports 123(2): 266-280.
[3] Kazdin (2000) “Encyclopedia of psychology”, Oxford University Press, New York.
[4] Kessler, Ronald C, Ayelet Meron Ruscio, Katherine Shear, and Hans-Ulrich Wittchen (2009) “Epidemiology of Anxiety Disorders”. Behavioral
Neurobiology of Anxiety and Its Treatment: 21-35.
[5] India State-Level Disease Burden Initiative Mental Disorders Collaborators. (2020) “The burden of mental disorders across the states of India:
the Global Burden of Disease Study 1990–2017.” The Lancet Psychiatry 7(2): 148-161
[6] Cho, Gyeongcheol, Jinyeong Yim , Younyoung Choi , Jungmin Ko, and Seoung-Hwan Lee (2019) “Review of Machine Learning Algorithms
for Diagnosing Mental Illness.” Psychiatry Investigation 16(4): 262-269.
[7] Srividya, M, S Mohanavalli, and N Bhalaji (2018) “Behavioral Modeling for Mental Health using Machine Learning Algorithms.” Journal
Of Medical Systems 42(5): 1-12.
[8] Wyatt, Tammy Jordan, Sara B. Oswalt and Yesenia Ochoa (2017). “Mental Health and Academic Success of First-Year College Students.
International Journal of Higher Education 6(3):178.
[9] Garcia-Ceja, Enrique, M. Riegler, T. Nordgreen, P. Jakobsen, K. Oedegaard, and J. Tørresen (2018) “Mental health monitoring with multimodal
sensing and machine learning: A survey.” Pervasive And Mobile Computing 51:1-26.
[10] Mohd Shafiee, Nor Safika, and Sofianita Mutalib (2020) “Prediction of Mental Health Problems among Higher Education Student Using
Machine Learning.” International Journal of Education and Management Engineering 10(6):1-9.
[11] Glaz, Aziliz le, Yannis Haralambous , Deok-Hee Kim-Dufor , Philippe Lenca , Romain Billot , Taylor C Ryan , Jonathan Marsh , Jordan
DeVylder , Michel Walter , Sofian Berrouiguet, and Christophe Lemey (2021) “Machine Learning and Natural Language Processing in Mental
Health: Systematic Review”. Journal of Medical Internet Research 23(5):1-20.
[12] Tiffin, Paul A, and Lewis W Paton. (2018) “Rise of the machines? Machine Learning Approaches and Mental Health: Opportunities and
challenges.” The British Journal of Psychiatry 213(3): 509–510.
[13] Dobias, Mallory L, Michael B Sugarman, Michael C Mullarkey, and Jessica L Schleider (2021) “Predicting mental health treatment access
among adolescents with elevated depressive symptoms: Machine learning approaches.” Administration and Policy in Mental Health and Mental
Health Services Research 49(1): 88–103.
[14] Liu, Yang S., J. Hankey, Bo Cao, and P. Chokka (2021) “Screening for major depressive disorder in a tertiary mental health centre using
Early Detect: A machine learning-based pilot study.” Journal of Affective Disorders Reports 3:100062.
[15] Abdullah, Azman, R. Abirami, A. Gitwina, and C. Varthana (2021) “Assessment of academic performance with the E-mental Health
Interventions in virtual learning environment using Machine Learning Techniques: A hybrid approach”. Journal of Engineering Education
Transformations 34:79-85.
[16] Albagmi, Faisal Mashel, Aisha Alansari, Deema Saad Al Shawan, Heba Yaagoub AlNujaidi, and Sunday O. Olatunji. (2022) “Prediction of
generalized anxiety levels during the Covid-19 pandemic: A machine learning-based modeling approach.” Informatics in Medicine Unlocked,
28:100854.
[17] Na, Kyoung-Sae, Seo-Eun Cho, and Seong-Jin Cho. (2021) “Machine learning-based discrimination of panic disorder from other anxiety
disorders.” Journal of Affective Disorders, 278:1–4.
[18] Bokma, Wicher A., Paul Zhutovsky, Erik J Giltay, Robert A Schoevers, Brenda W J H Penninx, Anton L J M van Balkom, Neeltje M Batelaan,
and Guido A van Wingen. (2020) “Predicting the naturalistic course in anxiety disorders using clinical and biological markers: A machine learning
approach.” Psychological Medicine, 52(1):57–67.
[19] Khullar, Vikas, Raj G. Tiwari, Ambuj K. Agarwal, and Soumi Dutta. (2021) “Physiological Signals Based anxiety detection using ensemble
machine learning”. Cyber Intelligence and Information Retrieval: 291: 597–608.
[20] Rakesh, Gopalkumar. (2017) “Suicide prediction with machine learning.” American Journal of Psychiatry Residents Journal, 12(1):15–17.
[21] Anbarasi, L. Jani, Malathy Jawahar, Vinayakumar Ravi, and Sherin M. Cherian. (2022) “Machine Learning Approach for anxiety and sleep
disorders analysis during COVID-19 lockdown”, Health and Technology 12(4): 825–838.