A

Synopsis Report
On
Predicting Sleep Disorders Using Machine Learning: A
Study Based On Demographic, Lifestyle & Psychological

Submitted in the partial fulfillment of the requirements

For award of the degree of

M.Sc in COMPUTER SCIENCE

(DATA SCIENCE AND ARTIFICIAL INTELLIGENCE)
(2023-2025)

Submitted To: - Submitted By :-

Dr. Vikram Singh Bharat Bhushan
Professor Roll No: 23111990872003
Department of CSE, CDLU M.Sc (DS & AI) 4th Sem.

Department of Computer Science & Engineering

Chaudhary Devi Lal University, Sirsa-125055
1. Introduction
Sleep is a critical biological function that plays a vital role in physical health, cognitive
performance, and emotional well-being. Poor sleep quality and sleep disorders, such as insomnia
and sleep apnea, have been linked to numerous adverse health outcomes, including
cardiovascular diseases, metabolic disorders, and mental health conditions The increasing
prevalence of sleep-related issues in modern society necessitates a deeper understanding of the
factors influencing sleep health. (Walker, 2017).
In recent years, the area where machine learning and mental health study meet has been very
popular. This is because it has the potential to completely change how we diagnose, treat, and
understand mental disorders. Using machine learning techniques is a hopeful way to improve our
ability to predict mental health conditions because technology is always changing. The point of
this study is to look into how machine learning can be used to find mental illnesses before they
happen. In the health field, machine learning (ML) is a type of artificial intelligence (AI) that
uses ways to teach computers how to learn. Deep learning (DL), guided learning, and
unsupervised learning are the three most common ways to learn about this subject. Two more
types of machine learning are reinforcement learning and semi-supervised learning.
Reinforcement learning uses algorithmic experiences and actions to learn by trial and error, and
semi-supervised learning mixes supervised and unsupervised learning (Graham et al., 2019).
Machine learning is an innovative approach that uses probability and data analysis to offer
customized therapies for mental health issues. By analyzing data from many sources, it can
identify indications and provide more effective treatment interventions. However, ethical data
use and transparent approaches must be addressed in machine learning research because of the
potential of skewed results and privacy issues. The overall objective of this project is to increase
our understanding of mental health and enhance clinical outcomes by examining the use of
machine learning techniques to predict sleep disorders. The goal of this study is to aid in the
creation of data-driven strategies for the early identification of sleep disorders.
There are three sections in the research proposal. The fundamental ideas of machine learning,
deep learning, and sleep disorder-related issues are covered in the first section, Introduction. The
literature on machine learning and deep learning-based methods for predicting sleep disorders is
thoroughly reviewed in the second section, "Literature Review," emphasizing the difficulties and
opportunities within. The study's goals and the suggested approach to achieving them are
covered in the section on research aims and methodology. It also talks about the research
instruments that will be utilized to carry out the study. It covers the research strategy, data
gathering techniques, and instruments used to carry out the investigation. This section clarifies
the execution of the study and the accomplishment of its goals by describing the research
method.
1.1 Machine learning
Machine learning is an area of artificial intelligence that focuses on creating algorithms and
models that enable computers to LEARN from (and make predictions or decisions) the data
without being explicitly programmed, in contrast to artificial intelligence, which focuses on
designing and developing computer-based systems that exhibit one or more traits of intelligent
behavior, such as listening, understanding, thinking, questioning, relating events, applying past
knowledge to new settings, abstraction, learning, creativity, etc. (fig. 1). Through experience, it
entails the study of statistical models and algorithms that enable computers to carry out tasks
automatically. Data patterns are found by machine learning algorithms, which then utilize them
to inform predictions or judgments. Natural language processing (NLP), picture and audio
recognition, recommendation systems, and predictive analytics are just a few of the fields in
which this technology finds use.

Fig.1 A schematic of AI, ML and DL

One must choose the specific method from the corpus of machine learning research and tools
used in order to create an usable machine learning model. Earlier and thorough understanding of
the issue domain, the availability of datasets, computational resources, and the duration of the
machine learning modeling effort are all necessary for this endeavor. Moreover, semi-structured,
unstructured, and structured datasets are all used by machine learning algorithms. A machine
learning system's overall effectiveness and efficiency, as well as the performance of the learning
algorithm, are influenced by the quality and quantity of the data. Data-driven systems in
classification analysis, regression, data clustering, feature engineering and dimensionality
reduction, association rule learning, and reinforcement learning can be rapidly constructed using
machine learning techniques. Moreover, deep learning originates from artificial neural networks,
a broader family of machine learning methods renowned for their ability to analyze data
intelligently. As a result, it could be challenging to choose a learning algorithm that works well
for a given application. This is because different learning algorithms have different objectives,
and depending on the characteristics of the data, different learning algorithms may yield different
outcomes even within a similar category. (Sarker, 2021b).
1.1.1 Taxonomy of Machine Learning
Machine learning as a discipline includes several learning types, different learning techniques
within each type, and potentially many algorithms within each machine learning method.
Gaining an understanding of machine learning's various applications and approaches requires an
understanding of its taxonomy (fig. 2).

Fig. 2 Taxonomy of Machine Learning

Supervised Learning
Supervised algorithms are machine learning algorithms that need outside assistance. The input
dataset is divided into train and test datasets. The training dataset's output variable must be
predicted or classified. Each algorithm employs patterns found in the training dataset to
categorize or predict data from the test datasets (Laurent, 2022).
Unsupervised Learning
When the training set's data is neither labeled nor classed, this kind of machine learning
technique is used. Systems that can deduce the explanation of a hidden component from
unlabeled data are the focus of unsupervised machine learning. Instead of assessing the right
answer, this kind of machine learning finds the information that allows an inference or deduction
to be made from the data to highlight a feature or hidden aspect from unlabeled data
(Hamarashid et al., 2022).
Semi-Supervised Learning
A specific kind of machine learning called semi-supervised learning uses both labeled and
unlabeled data to carry out learning tasks. It helps deal with the huge quantity of unlabeled data
seen in many use cases and is theoretically situated between supervised and unsupervised
learning. It is often used in conjunction with smaller collections of categorized data. A physical
trial for a learning challenge is also part of the labeled data gathering process. Therefore, while
processing unlabeled data is relatively cheap, the labeling procedure expenses make a branded
training set unfeasible. Semi-supervised learning may offer important functional advantages in
certain circumstances. Because it produces more precision with fewer human inputs, theory and
experience are both very valuable.
Reinforcement Learning
Through reinforcement learning, software agents and robots may automatically determine the
optimal behavior in a particular scenario to maximize performance. It involves the use of
exploration and exploitation methods to facilitate interaction between the learning agent and its
environment. While exploitation refers to the learning agent's actions based on information
gleaned from the environment, exploration refers to the learning agent's behavior through trial
and error. The environment uses moral behavior as a reinforcement cause, rewarding the learning
agent. The agent learns more about the environment and chooses the optimal course of action by
using the rewards it has acquired (Bansal, 2020).
1.1.2 Application Areas
Machine learning has recently found applications in several fields, transforming how activities
are completed and data insights are obtained. Machine learning has a significant and wide-
ranging impact on everything from healthcare to banking, retail to transportation. The various
application domains where machine learning methods and algorithms are being used to promote
efficiency and innovation are examined in this subsection.
Finance and banking
In order to enhance decision-making and lower financial risks, machine learning algorithms are
utilized in fraud detection, risk assessment, algorithmic trading, and customer relationship
management.
Natural Language Processing
By analyzing and synthesizing human language through machine learning, natural language
processing (NLP) applications enable sentiment analysis, chatbots, language translation, and
information extraction, all of which enhance engagement and interaction.
Healthcare and Medicine
Healthcare is undergoing a transformation because to machine learning, which provides
personalized patient monitoring, treatment planning, and diagnostics. The potential uses include
precision medicine, drug discovery, medical image analysis, and disease prediction.
E-commerce and Recommendation Systems
Recommendation engines are powered by machine learning, which enhances user experiences,
makes better product suggestions, and optimizes marketing tactics to increase consumer
engagement and sales.
Marketing and Advertising
ML algorithms promote campaign performance analysis, consumer segmentation, and targeted
advertising, which raises marketing ROI and customer engagement.
Environmental Sciences
ML supports biodiversity conservation, pollution monitoring, and climate modeling by shedding
light on environmental trends and encouraging sustainable practices.
 Cyber security
In order to improve cyber defenses and safeguard sensitive data, machine learning algorithms are
employed to secure networks, detect anomalies, and collect threat intelligence.
Transportation & Logistics
ML algorithms improve supply chain management, resource allocation, and route planning,
resulting in more cost-effective transportation systems and improved delivery procedures.
Education and E-learning
Machine learning improves individualized learning experiences, adaptive tutoring systems, and
educational content recommendations, addressing individual learning needs and enhancing
educational outcomes.

2. Literature Review
This section offers a thorough review of the literature with a focus on machine learning and deep
learning techniques used to forecast the sleep disorders. Along with summarizing the state of the
field's research, it will also point out potential and problems. This thorough analysis will provide
the groundwork for improving predictive models and guiding future lines of inquiry.
Al Sawah et al. (2015) conducted a cross-sectional study to examine how 98 first- and second-
year podiatric medical students' sleep quality was affected by their intake of coffee and energy
drinks as well as their perceptions of stress. The study found high levels of stress and poor sleep
quality using standardized measures like the Pittsburgh Sleep Quality Index (PSQI), Epworth
Sleepiness Scale, and Perceived Stress Scale (PSS-10). Logistic regression showed significant
relationships between caffeine consumption, stress, and poor sleep. Additionally, there was an
inverse relationship between academic achievement and poor sleep quality, indicating that
behavioral treatments may improve student outcomes and well-being.
Zheng et al. (2023) carried out a large-scale cross-sectional survey of 20,645 college students in
Fujian, China, during the COVID-19 pandemic to evaluate sleep quality with the Pittsburgh
Sleep Quality Index (PSQI). The authors used statistical and machine learning methods—such as
logistic regression, random forest, naive Bayes, and artificial neural networks (ANN)—to create
prediction models based on 11 most important variables like age, sex, caffeine intake, internet
use, and psychological stressors during the pandemic. Among the models, the ANN showed the
best accuracy (73.5%) and predictive power (AUC = 0.713), pointing to how behavioral and
emotional processes play a crucial role in sleep quality during public health emergencies.
In a cross-sectional study, Lund et al. (2010) examined sleep behavior and factors associated
with sleep problems in 1,125 college students from a Midwestern institution, ages 17 to 24.
Using recognized instruments such the Epworth Sleepiness Scale, the Pittsburgh Sleep Quality
Index (PSQI), and mood and stress scales, they evaluated sleep patterns and predictors of sleep
disturbances and found that over 60% of students slept poorly. With caffeine, alcohol, and sleep
schedule consistency not being significant predictors, multiple regression analysis revealed
psychological factors—most notably tension and subjective stress—as the greatest predictors of
sleep quality, accounting for 24% of the variance in PSQI scores.
Airlangga (2024) conducted a comprehensive comparative study using a dataset of 400
individuals that included demographic, lifestyle, and cardiovascular health metrics to predict
sleep disorders such as insomnia and sleep apnea. The study applied and evaluated multiple
machine learning models—including logistic regression, decision trees, SVM, neural networks,
and ensemble methods—using metrics like accuracy, precision, recall, and F1 score. Ensemble
techniques, particularly RandomForest and XGBClassifier, emerged as top performers with F1
scores around 0.93, demonstrating their robustness in managing complex, multidimensional
health data and their potential for clinical diagnostic use.
Alshammari (2024) used the Sleep Health and Lifestyle Dataset with 400 records and 13
features such as sleep duration, stress, BMI, and occupation to create a classification model for
sleep disorders like insomnia and sleep apnea. The research compared various machine learning
algorithms such as K-Nearest Neighbors, Support Vector Machine, Decision Tree, Random
Forest, and Artificial Neural Network (ANN), with and without hyperparameter tuning using
Genetic Algorithms. Findings indicated that the optimized ANN model with GA attained the best
classification accuracy (92.92%) with very good precision (92.01%), recall (93.80%), and F1-
score (91.93%), reaffirming its better ability to deal with intricate, multi-dimensional health data
in accurate sleep disorder prediction.
Rahman et al. (2025) performed a comprehensive study with the Sleep Health and Lifestyle
Dataset, which contains 374 instances with 13 demographic, behavioral, and physiological
attributes, to predict sleep disorders—None, Insomnia, and Sleep Apnea. The researchers
employed a solid machine learning pipeline consisting of data preprocessing, label encoding,
SMOTEENN for class imbalance handling, ANOVA for testing feature significance, and feature
engineering to enhance model inputs. The performance of fifteen machine learning algorithms
was tested, and the best performer was Gradient Boosting with an accuracy of 97.33%, precision,
recall, and F1-score of 0.9733, and AUC of 0.9953. The research showed that ensemble models'
higher performance and included feature selection and optimization as a priority in sleep disorder
classification
Anisha et al. (2024) presented an intelligent Chimp-based Recurrent Mets Framework (CbRMF)
for the classification and prediction of the severity of obstructive sleep apnea (OSA) from a
metabolism syndrome dataset of 769 patients. The model applied deep learning principles
coupled with a bio-inspired chimp optimization algorithm for efficient preprocessing, feature
extraction (e.g., BMI, glucose, blood pressure), and classification of OSA into moderate and
severe. When coded in Python, the CbRMF had a 99.1% accuracy, 96.6% recall, 95% F-score,
and a mere 0.8% error rate—better than standard ML models such as SVM, Random Forest, and
Naïve Bayes. The research showed the model's performance on different sets of data including
stroke unit and polysomnography data, validating its efficacy in real-world sleep disorder
prediction
The study (C et al., 2023) uses machine learning techniques for detecting stress, anxiety, and
depression. The study utilizes a dataset containing demographic information and scores related to
stress, anxiety, and depression. Notably, the proposed K-nearest neighbor (K-NN) model
achieved impressive accuracy, reaching 94.5% for stress and anxiety prediction, and 97.7% for
depression prediction. In contrast, the Naive Bayes model demonstrated slightly lower
accuracies: 81.8% for stress, 81.5% for anxiety, and 83.5% for depression. The dataset includes
responses to 21 questions about stress, anxiety, and depression, along with demographic data.
Moving forward, the study recommends exploring alternative machine learning algorithms for
affective disorder detection and incorporating additional demographic variables to enhance
prediction precision.
Authors in (Nemesure et al., 2021) investigate the application of machine learning to predict
major depressive disorder (MDD) and generalized anxiety disorder (GAD) using electronic
health records (EHRs). The study focuses exclusively on these two mental health conditions,
without addressing other psychological disorders or stress. Utilizing a dataset from 4,184
undergraduate students, the researchers excluded psychiatric features and 11 employed a
machine learning pipeline comprising six unique classifiers, including XGBoost, Random Forest,
 Support Vector Machine, K-nearest-neighbors, and a neural network. The model achieved an
AUC of 0.73 for predicting GAD and 0.67 for predicting MDD.
Researcher Kudari (2023) illuminate the critical topic of mental health challenges, with a
specific emphasis on depression, which is ranked as a major worldwide health concern by the
WHO. The survey carefully covers numerous machine learning approaches, such as
Convolutional Neural Networks (CNN), Artificial Neural Networks (ANN), Support Vector
Machines (SVM), and K-Nearest Neighbors (K-NN), used to predict various types of depression.
While the precise dataset is unknown, the work highlights the potential for combining deep
learning algorithms to handle complicated data with high-dimensional features, hence improving
mental health prediction. The survey emphasizes the importance of addressing computational
complexities and ethical considerations, advocating for further research to refine models, explore
additional features, and validate their efficacy in real-world scenarios, as well as incorporating
robust ethical and privacy safeguards into mental health prediction applications. This poll
emphasizes the revolutionary potential of machine learning and deep learning approaches for
improving mental health outcomes, as well as the significance of appropriate implementation.
Researchers Squires et al. (2023) explores the role of AI, ML, and DL in precision psychiatry
for depression. Methods include random forest classifiers for Twitter-based depression detection,
deep learning for perinatal depression diagnosis, and WeChat data utilization. Impressive
accuracies (91.3% with EEG-based K-NN, 85% with linear SVM) were achieved. However,
empirical validation and interdisciplinary collaboration are crucial. Addressing uncertainty and
model validation will enhance clinical implementation. collection methods, and tools utilized to
execute the study. By outlining the research process, this section lends clarity regarding the
study's execution and objectives’ achievement research tools slated to be used to conduct the
research. It includes the research approach, data.

3. Research Gap
There is a major research gap in applying machine learning (ML) for predicting and diagnosing
sleep disorders. There are a lot of studies out there, but most of them only look at a few specific
conditions. They don't look at things like stress, lifestyle, or other illnesses that may be present at
the same time. Also, there aren't many models that look at mental health from a wide range of
angles and use a single method. One of the biggest issues is that researchers usually build their
own datasets per study. This causes variations in how data is collected, tagged, and formatted, so
 comparing outcomes or building standard models is difficult. Consequently, data
representativeness and quality are therefore low, lowering the reliability of ML models in real
environments. Lastly, although both machine learning and deep learning have been applied
individually in individual studies, few have combined them within one model. Combining ML
and DL methods may improve accuracy, flexibility, and performance but is an area that has not
been fully explored. This gives a great opportunity for innovation and advancement in the use of
AI for sleep disorder.

4. Research Objectives and Methodology

Objectives of the study and explain how the research will be conducted and what research
method and tools will be included in the conduct of the research, it discusses the methodology,
data collection methods, and tools used to implement the study. Describing the research process,
this part is transparent and clear about the implementation and objectives of the study.
Statement
4.1 Research Objectives
I. To study the techniques used for sleep disorders using machine and deep learning.
II. To preprocess the dataset used for sleep disorders.
III. To propose the machine learning model for sleep disorder predictions.
IV. To evaluate the performance of the proposed model and compared with the existing
techniques.
4.2 Research Methodology
I. Data collection: Gather information from several repositories and prepare the data
according to the requirements.
II. Model design: Create ML and DL models for early detection, choose effective
algorithms, and combine data from various sources.
III. Model Evaluation: Use resources to obtain performance and implement cross-
validation methods.
IV. Comparative analysis: Compare the efficacy of ML and DL models.
Examine about efficiency and complexity.
V. Statistical Interpretation: Use statistical tools to examine the results.
5. References
Airlangga, G. (2024). Evaluating Machine Learning Models for Predicting Sleep Disorders in a
Lifestyle and Health Data Context. JIKO (Jurnal Informatika dan Komputer), 7(1), 51–
57. https://doi.org/10.33387/jiko.v7i1.7870
Al Sawah, M., Ruffin, N., Rimawi, M., Concerto, C., Aguglia, E., Chusid, E., Infortuna, C., &
Battaglia, F. (2015). Perceived Stress and Coffee and Energy Drink Consumption
Predict Poor Sleep Quality in Podiatric Medical Students: A Cross-sectional Study.
Journal of the American Podiatric Medical Association, 105(5), 429–434.
https://doi.org/10.7547/0003-0538-105.5.429
Alshammari, T. S. (2024). Applying Machine Learning Algorithms for the Classification of
Sleep Disorders. IEEE Access, 12, 36110–36121.
https://doi.org/10.1109/ACCESS.2024.3374408

Anisha, P. R., Reddy, C. K. K., Hanafiah, M. M., Murthy, B. R., Mohana, R. M., & Pragathi, Y.
V. S. S. (2024). An intelligent deep feature based metabolism syndrome prediction
system for sleep disorder diseases. Multimedia Tools and Applications, 83, 51267–
51290. . https://doi.org/10.1007/s11042-023-17296-4
Bansal, H., & Sharma, K. (2020). A review study on various algorithms of machine learning.
Journal of emerging technologies and innovative research.
C, K. K. R., Tungana, B., Hanafiah, M. M., & Doss, S. (2023). A study of Machine Learning
Methods Based Affective Disorder Detection using Multi-Class Classification
[Preprint]. In Review. https://doi.org/10.21203/rs.3.rs-2888288/v1
Graham, S., Depp, C., Lee, E. E., Nebeker, C., Tu, X., Kim, H.-C., & Jeste, D. V. (2019).
Artificial Intelligence for Mental Health and Mental Illnesses: An Overview. Current
Psychiatry Reports, 21(11), 116. https://doi.org/10.1007/s11920-019-1094-0
Hamarashid, H., Qader, S. M., Saeed, S. A., Hassan, B. A., & Ali, N. A. (2022). Machine
Learning Algorithms Evaluation Methods by Utilizing R. UKH Journal of Science and
Engineering, 6(1), 1–11. https://doi.org/10.25079/ukhjse.v6n1y2022.pp1-11
Kudari, J. M. (2023). Detection And Prediction Of Mental Health Illness Using Machine
Learning And Deep Learning Techniques: A Survey. 11(12).
Liu, K., Droncheff, B., & Warren, S. L. (2022). Predictive utility of symptom measures in
classifying anxiety and depression: A machine-learning approach. Psychiatry Research,
312, 114534. https://doi.org/10.1016/j.psychres.2022.114534

Lund, H. G., Reider, B. D., Whiting, A. B., & Prichard, J. R. (2010). Sleep Patterns and
Predictors of Disturbed Sleep in a Large Population of College Students. Journal of
Adolescent Health, 46(2), 124–132. https://doi.org/10.1016/j.jadohealth.2009.06.016
Nemesure, M. D., Heinz, M. V., Huang, R., & Jacobson, N. C. (2021). Predictive modeling of
depression and anxiety using electronic health records and a novel machine learning
approach with artificial intelligence. Scientific Reports, 11(1), 1980.
https://doi.org/10.1038/s41598-021-81368-4
Rahman, M. A., Jahan, I., Islam, M., Jabid, T., Ali, M. S., Rashid, M. R. A., & Ali, M. A. (2025).
Improving Sleep Disorder Diagnosis through Optimized Machine Learning Approaches.
IEEE Access. https://doi.org/10.1109/ACCESS.2025.3535535

Sarker, I. H. (2021b). Machine Learning: Algorithms, Real-World Applications and Research

Directions. SN Computer Science, 2(3), 160. https://doi.org/10.1007/s42979-021-00592-
x
Walker, M. (2017). Why we sleep: Unlocking the power of sleep and dreams. Simon and
Schuster.
Zheng, W., Chen, Q., Yao, L., Zhuang, J., Huang, J., Hu, Y., & Wang, Y. (2023). Prediction
models for sleep quality among college students during the COVID-19 outbreak: cross-
sectional study based on the internet new media. Journal of Medical Internet Research,
25, e45721. https://doi.org/10.2196/45721