International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 09 | Sep 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 319
A Machine Learning Perspective on Emotional Dichotomy during the
Pandemic
Dr.A.Sharada
1
, Manasa Jonnalagadda
2
1
Professor, Computer Science & Engineering, G Narayanamma Institute of Technology & Science, Telangana, India
2
M.Tech Student, Computer Science & Engineering, G Narayanamma Institute of Technology & Science,
Telangana, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Mental health is a stabilizing force of an individual’s emotional well-being, and any distress can cause imbalances in
one’s conventional routine and plethora of mental disorders. Mental health concerns usually took a backseat during the pandemic
and impacts seamless functioning for teachers and students in educational environment. Depression is a mental health condition
manifesting constant elevation or lowering of person’s mood and little interest in everyday activities causing substantial
impairment in everyday life. Depression in particular is influenced by complex array of factors including everyday stress, academic
strain, compounded negative emotions and panic due to COVID-19 outbreak. Research conducted in healthcare domain in par with
Artificial Intelligence provides various methods for detection and diagnosis of depression. However, minimal research is conducted
predicting depression based on individual’s situation and their environment in early stages. The objective of this study is to propose
a context aware model for teachers and students for predicting risk of depression in educational framework and pandemic. The
datasets are created through structured self-reporting questionnaires and potential variables for depression risk are identified
with Regression analysis. Related context information is extracted in relevance with each potential variable and Convolutional
Neural Networks is applied for depression risk prediction. Subsequently, accuracy of the proposed model for teachers and students
is evaluated with performance metrics and comparative analysis of Multiple Regression and Convolutional Neural Networks.
Key Words: Mental Health, Depression Risk, Convolutional Neural Networks, Multiple Regression, Machine Learning.
1. INTRODUCTION
COVID-19 is a global humanitarian cataclysm that has left the world in shambles over the recent years. In India, it has enforced
rapid transition in education, IT, healthcare, and other sectors, to digitize and implement various strategies for their seamless
functioning [10]. Specifically, schools and colleges were forced to run emergency online learning/classes causing prolonged
social isolation and increasing academic stressors on both teachers and students. It also had major impact on everyone’s life,
disturbing individual’s conventional activities along with their physical and mental health. Mass fear and uncertainty has
reflected disparaging effect in holistic well-being of a person steering strong emotions like stress, anxiety, anger, depression,
and other complex array of factors. Work stress, difficult financial situation, family issues, personal and professional problems,
changes due to the COVID-19 and other psychological and environmental parameters originating from an individual’s way of
life contribute to distress and mental health disorders.
According to World Health Organization (WHO), depression ranks high among common mental debilities. Depression is a
mental health condition manifesting constant elevation or lowering of person’s frame of mind and loss of interest in daily
activities causing substantial impairment in everyday life. Given the current shifts in the educational landscape over the past
years, depression has become increasingly common in teachers and students in India. It is an emotional dichotomy found in
various strata of the society and in different age groups. Parameters like complete burnout, extreme work strain due to
academic and curricular responsibilities in teachers; and academic stressors, peer and societal pressure in students could afflict
the individual’s ecosystem. Thus, a souring need rises to support the emotional well -being of teachers and students by
predicting depression risk in preliminary stages to potentially reduce the escalation of the illness and in turn improve their
quality of life.
Machine Learning and Deep Learning based mental health explorations [11] have attracted lot of attention to predict mental
disorders using multimodal data like text, images, and videos. Approaches like Deep Neural Networks (DNN) and Regression
has opened a new frontier to address early screening, detection, prediction, and diagnosis of various disorders by tracking
compound emotional parameters associated with the mental health challenges. The statistical and computational methods
extended by Machine Learning assist in constructing robust automated prediction and detection of depressive symptoms with
the ability to learn and train from data. Multimodal data relying on frequent measurements of depression status procured from
various sources have been implemented with deep learning models for early recognition of depression symptoms in the
individuals. However, minimal research exists for classifying and predicting individual’s emotional state based on their