Step-by-Step Final Project Report (FPR) Writing Guide by Crowjack

The abstract provides a concise and impactful summary of your project. It's a short, powerful summary that gives the reader a quick overview of your entire project. Think of it as the first (and sometimes only) thing someone will read to decide if your report is worth their time. Since you have a limit of half a page, every sentence needs to count.

What to Include in Your Abstract?

Your abstract should address these key points:

• The Aim (What are you trying to achieve?) - Start with a clear statement of your project's purpose. What problem are you trying to solve, or what question are you trying to answer?
• The Methodology (How are you going about it?) - Briefly describe your research methods. What approach are you using to investigate your research question? If you're using a quantitative approach, mention the type of data you're collecting and how you're analysing it.
• Potential Outcomes (What do you expect to find?) - Outline the expected results or potential conclusions. What do you anticipate discovering through your research? This can be framed as your hypothesis or expected trends.

Example of Abstract

The study developed a non-intrusive voice analysis- based system for the early detection of depression by examining vocal features such as pitch, tone, and speaking patterns through machine learning techniques. Utilizing established datasets like DAIC-WOZ and TESS, the research gathered and pre-processed voice recordings to extract significant features that correlate with depressive states. Employing machine learning models, specifically Support Vector Machines (SVM) and Long Short-Term Memory (LSTM) networks, the study evaluated their performance in detecting depression based on metrics such as accuracy, precision, recall, and F1 score. The methodology included a quantitative research design that involved an extensive literature review to identify research gaps, followed by experimental testing on existing speech datasets. Results indicated that the LSTM network achieved an accuracy of 85% and an F1 score of 0.82, outperforming the SVM model, which achieved 78% accuracy and an F1 score of 0.75. These outcomes suggest enhanced diagnostic accuracy and early intervention capabilities in mental health care by providing a novel tool that improves upon traditional assessment methods reliant on subjective evaluations. Eventually, this research demonstrated how machine learning approaches can facilitate the recognition of depressive patterns within speech, thereby contributing to more effective mental health diagnostics.

The introduction sets the stage for your entire project. It's your opportunity to grab the reader's attention, provide necessary background information, and clearly state the purpose and scope of your research. Think of it as a roadmap that guides the reader through the rest of your report.

Key Components of Your Introduction

Your introduction should address the following elements, flowing logically from one to the next:

1. Background:

• The background section of your introduction is crucial for setting the stage for your research. It introduces your research area to the reader, providing context and justifying the need for your study.
• Start broad, introducing the reader to the general subject area. The key is to begin with something easily understandable and relevant to your topic.
  • For example, if your dissertation focuses on the topic “Voice-Based Detection of Depression Using Machine Learning Techniques”; given the rise in mental health awareness and the increasing prevalence of depression, there’s a growing need for accessible and objective methods for early detection (Foulkes and Andrews, 2023). Depression is a significant global health concert, affecting millions worldwide and impacting various aspects of life (WHO, 2024).
• Gradually narrow the focus from this broad context to the specific area your research addresses. This involves introducing key concepts, trends, or debates within your broader topic.
  • In the voice-based depression detection example, you could explain that traditional methods of diagnosing depression rely on self-reported questionnaires and clinical interviews, which can be subjective and time-consuming. This has led to the exploration of more objective and efficient methods, such as voice analysis, which leverages machine learning to identify vocal biomarkers of depression (Richter et al., 2021).
• It is then helpful to provide relevant background information to help your reader understand the current state of knowledge in your area. This might include historical context, key theories or models, or important statistics or trends.
• Make a direct and clear link between the background information and your specific research topic. Show how your research fits into the existing literature and why it's a logical next step.
  • For example, you could state that research in voice-based depression detection has explored various acoustic features, such as pitch, tone, and speech rate, and has employed machine learning algorithms like Support Vector Machines (SVM) and Long Short-Term Memory (LSTM) networks to classify individuals as depressed or non depressed (Yalamanchili et al., 2020).
  • This sentence is crucial because it bridges the gap between the general background and your specific research question.
• The final part of your background section should naturally lead into the problem statement. This is where you highlight the gap in knowledge, the unanswered question, or the issue that your research will address.
  • Think of it as setting up the "problem" that your dissertation will "solve."
  • This lack of understanding is problematic because it limits the ability to develop robust and generalizable voice-based depression detection systems that can be deployed in real-world settings.

2. Aim:

Write a brief statement that outlines the primary goal of your research. The aim of your research should clearly articulate the purpose of your project in a single statement, ideally beginning with an infinitive verb (e.g., “to determine, “to investigate”, “to develop”), followed by an action that captures the essence of your inquiry. This ensures that the aim addresses what you intend to investigate, how you plan to approach it, and remains concise while guiding your research direction.

For example: This study aims to investigate the application of a non-intrusive voice analysis-based system to detect the early signs of depression by analyzing voice features such as pitch, tone, and speaking patterns with the aid of machine learning methods.

3. Objectives:

When crafting objectives for your experiment or research project, it’s important to articulate specific goals that detail what you intend to achieve. Each objective should begin with the word “to”, followed by an action verb that clearly defines the task at hand. Examples of effective action verbs include “to review”, “to analyse”, “to design”, “to compare”, “to measure”, and “to evaluate”. This structure helps in outlining the steps needed to fulfil your aim and provides clarity on the direction of your study.

For example:

• To gather and preprocess voice recordings from established depression-related datasets.
• To identify and extract important voice features which correspond to depressive states.
• To implement machine learning models-SVM, LSTM-and evaluate performance in detecting depression using voice data.
• To investigate comparison between voice analysis and standard practice for detection of depression.

To ensure your objectives are well-defined and achievable, it's helpful to use the SMART framework. SMART stands for:

S: Specific - What Exactly Do You Want to Achieve?

In order for an objective to be effective, it needs to be specific. A specific objective answers question like:

• What needs to be accomplished?
• Who is responsible for it? (In the context of a student project, this is usually you.)
• What specific steps need to be taken to achieve it?

Thinking through these questions helps get to the heart of what you’re aiming for.

M: Measurable - How Will You Track Progress?

Specificity is a solid start, but quantifying your objectives (that is, making sure they’re measurable) makes it easier to track progress and know when you’ve reached the finish line. To make an objective measurable, ask:

• How will I measure progress towards this objective?
• What specific metrics will I use?
• What does success look like?

A: Achievable - Is This Objective Realistic?

This is the point in the process when you give yourself a serious reality check. Objectives should be realistic, i.e., not setting yourself up for failure. Ask yourself 🡪

• Is this objective something I can reasonably accomplish given my resources, skills, and time constraints?
• Am I being overly ambitious?
• What potential obstacles might I face, and how can I mitigate them?

R: Relevant - Why Is This Objective Important?

Here’s where you need to think about the big picture. Why are you setting the objective that you’re setting? How does it connect to your overall research question or project aim? Ask yourself 🡪

• How does this objective contribute to the overall goals of the project?
• Is this objective essential, or just a "nice-to-have"?
• Will achieving this objective make a meaningful contribution to the field (even if small)?

T: Time-bound - When Will You Achieve This Objective?

To properly measure success, you need to be clear about when an objective has been reached. What’s your time horizon? When will you start working on the tasks required to achieve the objective? When will you finish? SMART objectives should have time-related parameters built in, so you stay on track.

Example:

4. Research Question(s):

• To formulate effective research questions, first identify the specific issue or problem that your research will address.
• Each question should begin with interrogative words such as "what," "how," "why," or "to what extent."
• Ensure that the questions are clear and focused; they should be neither too broad nor too narrow, analytical rather than descriptive, and framed to require in-depth exploration rather than simple yes-or-no answers.
• Additionally, make sure that the questions align with your study’s purpose and include sufficient material for research while remaining within the scope of your study.

Example:

RQ1: How can machine learning approaches make it easier to recognize depressive patterns within speech?

5. Significance of the Research:

• Begin by explicitly stating the core problem or knowledge gap your research addresses. Highlight how your study offers solutions to the problems identified.
• Clearly describe the potential contributions and implications of your project to the field, emphasizing its importance, timeliness, and relevance.
• Explain what new insights your research provides, indicating whether it challenges existing theories, refines methodologies, or explores understudied phenomena.
• Specify who benefits from your findings, whether academics, practitioners, policymakers, or the broader society, and detail how they benefit.
• Substantiate your claims by referring to relevant literature and gaps that your study aims to fill.
• Highlight any novel aspects, like unique perspectives or innovative approaches.
• By directly and concisely demonstrating the theoretical, practical, methodological, or societal impact of your work, you convincingly establish its significance and justify its pursuit. Remember to remain focused and ensure your claims are well-supported.

Example:

This research holds significant potential for advancing the field of mental health by offering a non-intrusive, voice-based system for the early detection of depression (Shin and Bae, 2024). By leveraging machine learning techniques to analyze vocal characteristics, this study aims to provide a more objective and accessible method for identifying depressive symptoms compared to traditional diagnostic approaches that rely heavily on subjective self-assessments and clinical evaluations. The successful implementation of this system could lead to earlier interventions, improved treatment outcomes, and reduced burden on healthcare professionals (Habtamu et al., 2023). Further, the findings from this research can contribute to a deeper understanding of the relationship between vocal patterns and mental states, potentially informing the development of more sophisticated diagnostic tools and personalized treatment strategies in the future. Eventually, this work seeks to bridge the gap between technological innovation and mental health care, making early detection and intervention more readily available to individuals in need, regardless of their geographical location or access to specialized healthcare services.

6. Structure of the Dissertation

• This section provides an overview of how the dissertation is organized.
• Briefly describe each chapter or major section and explain its purpose within the overall research project.
• Start with the introduction, highlighting its role in setting the stage and presenting the research question.
• Then, summarize the literature review, explaining how it provides context and identifies gaps in existing knowledge.
• Next, outline the methodology chapter, emphasizing its description of the research design and data collection methods.
• Briefly explain the data analysis and results chapters, noting how they present the findings of the study.
• Finally, summarize the discussion or conclusion chapter, explaining how it interprets the results and relates them back to the research question.
• By providing this roadmap, you help the reader understand the flow of your dissertation and the contribution of each part.

Example:

• Abstract: A concise summary of the dissertation's purpose, methods, key findings, and conclusions.
• Chapter 1: Introduction: Presents the research problem, background, objectives, research questions, and significance of the study.
• Chapter 2: Literature Review: Critically analyses existing research related to voice-based depression detection and identifies gaps in the current knowledge.
• Chapter 3: Methodology: Describes the research design, data collection methods, feature extraction techniques, and machine learning models used for voice analysis.
• Chapter 4: Results: Presents the findings of the machine learning models' performance in detecting depression using voice data, including accuracy, precision, recall, and F1-score.
• Chapter 5: Discussion and Conclusion: Interprets the results, discusses their implications, compares the voice analysis method with standard practices, highlights limitations, and suggests directions for future research.
• References: Lists all the sources cited in the dissertation.

References

• Fenza, G., Gallo, M., Loia, V., Gaeta, M. and Herrera-Viedma, E. (2021). Data set quality in Machine Learning: Consistency measure based on Group Decision Making. 106, pp.107366–107366. doi:https://doi.org/10.1016/j.asoc.2021.107366.
• Foulkes, L. and Andrews, J.L. (2023). Are mental health awareness efforts contributing to the rise in reported mental health problems? A call to test the prevalence inflation hypothesis. New Ideas in Psychology, [online] 69(1), p.101010. doi:https://doi.org/10.1016/j.newideapsych.2023.101010.
• Habtamu, K., Birhane, R., Demissie, M. and Fekadu, A. (2023). Interventions to improve the detection of depression in primary healthcare: systematic review. Systematic Reviews, 12(1). doi:https://doi.org/10.1186/s13643-023-02177-6.
• Richter, T., Fishbain, B., Richter-Levin, G. and Okon-Singer, H. (2021). Machine Learning-Based Behavioral Diagnostic Tools for Depression: Advances, Challenges, and Future Directions. Journal of Personalized Medicine, 11(10), p.957. doi:https://doi.org/10.3390/jpm11100957.
• Shin, J. and Bae, S.M. (2024). Use of voice features from smartphones for monitoring depressive disorders: Scoping review. Digital Health, [online] 10. doi:https://doi.org/10.1177/20552076241261920.
• WHO (2024). Depression. [online] World Health Organization. Available at: https://www.who.int/health-topics/depression#tab=tab_1.
• Yalamanchili, B., Kota, N.S., Abbaraju, M.S., Nadella, V.S.S. and Alluri, S.V. (2020). Real-time Acoustic based Depression Detection using Machine Learning Techniques. [online] IEEE Xplore. doi:https://doi.org/10.1109/ic-ETITE47903.2020.394.

A literature review is a crucial component of academic research that serves several important purposes (Snyder, 2019). Below are key concepts and explanations to help you understand what a literature review involves and why it is essential:

Definition of a Literature Review:

A literature review is a comprehensive survey of existing research related to a specific topic or question. It involves critically examining and synthesizing previous studies, theories, and findings in order to provide a foundation for your own research.

Purpose of Conducting a Literature Review:

• Identify Gaps in Knowledge: A literature review helps identify gaps in the current body of knowledge, highlighting areas where further research is needed (Bezet, 2023).
• Establish Context: It provides context for your research by situating it within the existing literature, allowing you to demonstrate how your work builds upon or diverges from previous studies.
• Support Your Research Design: By reviewing existing methodologies and findings, you can refine your own research design and approach.
• Avoid Duplication: Conducting a thorough literature review ensures that you are not duplicating previous studies, allowing you to contribute new insights to the field.

Most universities emphasize the importance of conducting a comparative review during the literature review process, as it provides a more in-depth analysis than a summative review.

Overview of Comparative Literature Review:

• A comparative review analyses and contrasts different studies, theories, or methodologies within your identified themes, leading to a deeper understanding and highlighting gaps. There are two primary approaches to conduct this:
• This approach focuses on comparing and contrasting different studies, theories, or methodologies within your field.
• It allows researchers to analyse how various pieces of research relate to one another, identifying similarities, differences, and trends (Lai and Fong, 2024).

Importance of Comparative Reviews:

• A comparative review not only highlights what has been done in the field but also critically evaluates the strengths and weaknesses of existing research.
• It helps in understanding how different studies contribute to the broader discourse on a topic and can reveal conflicting findings that require further investigation (Marjan, 2017).

Process to Write Literature Review

Introduction to the Chapter:

• Clearly state the purpose and scope of the literature review.
• Outline the key themes or areas that will be explored.
• Research Sources: Specify the databases and search engines used to locate relevant literature (e.g., ResearchGate, ScienceDirect, MDPI, Google Scholar, Web of Science, Scopus). Justify the selection of these sources.
• Search String: Provide the exact search string(s) used to identify relevant articles (e.g., "(cybersecurity AND machine learning) OR (intrusion detection AND AI)").
• Keywords: List the key words used in the search process (e.g., cybersecurity, machine learning, intrusion detection, artificial intelligence). Explain why these keywords were chosen.

Example of Introduction to the Chapter:

This literature review chapter investigates the field of voice analysis for depression detection, examining its potential to augment and refine traditional diagnostic methods. The review will explore key themes, including the use of vocal characteristics as biomarkers, the application of machine learning models in analysing voice data, and the integration of voice analysis in clinical settings. The primary sources utilized include academic databases such as Google Scholar, ScienceDirect, and PubMed, chosen for their comprehensive coverage of medical and technological research. The search process employed keywords such as "voice analysis," "depression detection," "machine learning," "vocal features," and "acoustic analysis," along with search strings combining these terms to identify relevant studies and seminal works in the field.

((((((((voice analysis) AND (depression detection)) OR (acoustic analysis)) AND (depression)) OR (vocal features)) AND (depression)) AND (machine learning)) OR (speech analysis)) AND (mental health)

The aim is to provide a robust overview of current knowledge and identify research gaps to inform future research.

Literature Review

Theme Development - Creating themes for your literature review is an essential step in organizing your research and presenting your findings in a coherent manner (Doheny, 2023). In order to craft the themes, a step by step guide given below will help you develop effective themes:

Step 1: Identify Key Topics:

• Begin by reviewing your research questions and objectives. What are the main areas of focus in your study?
• Look for recurring topics or concepts in the literature you have gathered. These will serve as the foundation for your themes.

Step 2: Group Related Studies

• As you read through the literature, group studies that address similar aspects of your research question. This could include studies that use similar methodologies, address similar populations, or explore related findings.
• Consider creating a spreadsheet or document to track these studies and their key points, which will help you visualize how they connect.

Step 3: Develop Descriptive Titles

• For each group of related studies, create a descriptive title that accurately reflects the content of that section. Avoid generic labels like "Theme 1" or "Theme 2." Instead, use titles that convey the specific focus of the theme.
• Example:
  • Instead of "Theme 1," use "Overview of Voice Analysis in Depression Detection."
  • Instead of “Theme 2,” use “Current Machine Learning Techniques for Voice-Based Depression Assessment.”
  • Instead of “Theme 3,” use “Challenges and Limitations of Voice-Based Depression Detection Systems.”

After developing the themes, the next step is to conduct comparative literature review. A comparative review analyses and contrasts different studies, theories, or methodologies within your identified themes, leading to a deeper understanding and highlighting gaps. There are two primary approaches to conduct this:

1. Group Study Approach:

• How it Works: After identifying your themes, you group studies that address similar aspects of your research question.
  • This is useful when you've already gathered a collection of papers through initial searches and want to organize and synthesize them.
• 2. Manual, Theme-Driven Search:

How it Works: Instead of relying on pre-existing groups of studies, you search for papers theme by theme, actively seeking sources to complete each section. This is useful when you want to ensure a comprehensive overview of the literature related to each theme.

Steps for Conducting a Comparative Review (Applies to Both Approaches):

Summarize and Critically Evaluate:

For each theme, do NOT simply summarize one paper at a time. Instead:

• Craft a sentence or two presenting a key idea related to the theme.
• Immediately support that idea by referencing multiple authors who have also found similar results or proposed similar theories. This shows a synthesis of ideas, not just a summary of one paper.
• Example: "Machine learning models have shown promise in detecting depression through voice analysis (Sun et al., 2023). In this regard, (Ahmed et al., 2022) exemplified that these models often focus on acoustic features like pitch and speech rate."

Incorporate Contrasting Views:

• If you want to present a negative or contrasting viewpoint, use transition words or phrases like "in contrast," "however," or "conversely" to signal that you are presenting an opposing argument or limitation.
• Example: "Machine learning models have shown promise... However, these models often struggle with generalization across diverse populations (Ganatra, 2025). In contrast, (Mahmood et al., 2025) conducted a study and stated that traditional methods offer more reliable results..."

Identify Key Findings and Debates:

• Within each theme, highlight areas of agreement and disagreement. What are the major debates in the field? What questions remain unanswered?
• Example: "While there is agreement on the utility of acoustic features, the optimal machine learning algorithm for voice-based depression detection remains a topic of debate. SVM and LSTM models are widely used, but..."

Cite Relevant Sources:

• Use a consistent citation style (APA, Harvard).
• Potential Sources for Papers:
  • Google Scholar
  • Scopus
  • PubMed
  • IEEE Xplore
  • ScienceDirect
  • Web of Science
  • JSTOR
  • ResearchGate

Key Rules to Remember:

• No Single-Paper Summaries: Avoid summarizing one journal paper at a time. Integrate information from multiple sources into each paragraph.
• Support with Evidence: Every statement should be supported by evidence from the literature (i.e., citations).
• Acknowledge Contrasting Views: Use transition words to introduce opposing arguments or limitations.

Example of Literature Review:

Theme 1: Overview of Voice Analysis in Depression Detection : As per (Leal, Ntalampiras and Sassi, 2024), voice analysis has emerged as a promising tool in the detection of depression, leveraging acoustic features to provide objective assessments that complement traditional diagnostic methods. In this regard, (Ellgring and Scherer, 2022) conducted a study and stated that various vocal characteristics, such as pitch, tone, and speech rate, can serve as indicators of depressive states. For instance, individuals with depression often exhibit lower vocal volume and slower speech rates, which can be quantitatively analysed using machine learning algorithms (Ellgring and Scherer, 2022). Further to this, (Caroline Wanderley Espinola et al., 2022) demonstrated that these algorithms can identify patterns in voice recordings that correlate with depression severity, thus facilitating early detection and intervention. In addition, (Low, Bentley and Ghosh, 2020) conducted a study and mentioned that recent advancements in artificial intelligence have further enhanced the potential of voice analysis in mental health diagnostics. In this context, (Huang et al., 2024) described that models like wav2vec 2.0 have been utilized to extract high-quality voice features from audio data, significantly improving classification accuracy for depression detection. This approach allows for the identification of subtle changes in vocal patterns associated with mood disorders, providing clinicians with a powerful tool for monitoring patient progress over time (Huang et al., 2024). Further, (Wang et al., 2023) explained that changes in specific acoustic features can reflect treatment responses, such as improvements following cognitive-behavioural therapy. Apart from this, (Gawali et al., 2025) conducted a study and highlighted that the integration of voice analysis into clinical practice offers a low-cost and scalable solution for mental health screening. By utilizing voice as a biomarker for depression, healthcare providers can enhance their diagnostic capabilities and tailor treatment plans more effectively (Gawali et al., 2025). Lastly, (Saraswati and Andi Wirawan, 2024) specified that as research continues to validate these methods, voice analysis could play a critical role in transforming how depression is detected and managed, eventually leading to better patient outcomes through timely and accurate interventions.

3. Research Gap:

• Clearly identify the gap(s) in the existing literature that your research aims to address.
• Explain the significance of this gap and why it is important to address it.
• Describe how your research will contribute to filling this gap and advancing knowledge in the field.
• Specifically explain how your research design and methodology are suited to address the identified gap

Example of Research Gap:

The existing literature highlights significant advancements in using voice analysis for detecting depression, yet several research gaps remain unaddressed. Notably, while studies have demonstrated the effectiveness of various machine learning models in analysing vocal characteristics, there is a lack of comprehensive investigations that employ a non-intrusive voice analysis system specifically aimed at early detection of depression. Further, many current approaches rely on limited datasets, which may not adequately capture the diversity of vocal features associated with different depressive states. According to the (Institute for Health Metrics and Evaluation, 2021), depression affects an estimated 13.9% of the global population, highlighting the urgent need for effective detection methods. This study aims to fill these gaps by systematically gathering and pre-processing voice recordings from established depression-related datasets, extracting critical voice features linked to depressive states, and implementing robust machine learning models such as Support Vector Machines (SVM) and Long Short-Term Memory (LSTM) networks. As well, it will compare the efficacy of voice analysis against traditional diagnostic practices, thereby enhancing the understanding of how voice features can serve as reliable biomarkers for depression detection and monitoring treatment responses.

References

• Caroline Wanderley Espinola, Juliana Carneiro Gomes, Pereira, S. and Santos (2022). Detection of major depressive disorder, bipolar disorder, schizophrenia and generalized anxiety disorder using vocal acoustic analysis and machine learning: an exploratory study. Research on Biomedical Engineering, 38(3), pp.813–829. doi:https://doi.org/10.1007/s42600-022-00222-2.
• Ellgring, H. and Scherer, K.R. (2022). Vocal indicators of mood change in depression. Journal of Nonverbal Behavior, 20(2), pp.83–110. doi:https://doi.org/10.1007/bf02253071.
• Gawali, A., Gawali, C., Babhulkar, N., Shastrakar, M. and Zamare, G. (2025). A Project Report on Voice Analysis for Disease Screening. [online] Available at: https://www.ssgmce.ac.in/uploads/UG_Projects/cse/202324/Project%20Report%20Gr.%20No.%2013_2023-24.pdf [Accessed 14 Feb. 2025].
• Huang, X., Wang, F., Gao, Y., Liao, Y., Zhang, W., Zhang, L. and Xu, Z. (2024). Depression recognition using voice-based pre-training model. Scientific Reports, 14(1). doi:https://doi.org/10.1038/s41598-024-63556-0.
• Leal, S.S., Ntalampiras, S. and Sassi, R. (2024). Speech-based Depression Assessment: A Comprehensive Survey. IEEE Transactions on Affective Computing, pp.1–16. doi:https://doi.org/10.1109/taffc.2024.3521327.
• Low, D.M., Bentley, K.H. and Ghosh, S.S. (2020). Automated assessment of psychiatric disorders using speech: A systematic review. Laryngoscope Investigative Otolaryngology, 5(1), pp.96–116. doi:https://doi.org/10.1002/lio2.354.
• Saraswati, R. and Andi Wirawan (2024). From Words to Well-being: Analyzing Systems for Depression Detection through Speech. Asian American Research Letters Journal, [online] 1(4). Available at: https://aarlj.com/index.php/AARLJ/article/view/67 [Accessed 14 Feb. 2025].
• Wang, Y., Liang, L., Zhang, Z., Xu, X., Liu, R., Fang, H., Zhang, R., Wei, Y., Liu, Z., Zhu, R., Zhang, X. and Wang, F. (2023). Fast and accurate assessment of depression based on voice acoustic features: a cross-sectional and longitudinal study. Frontiers in Psychiatry, [online] 14. doi:https://doi.org/10.3389/fpsyt.2023.1195276.
• Ahmed, A., Aziz, S., Toro, C.T., Alzubaidi, M., Irshaidat, S., Serhan, H.A., Abd-alrazaq, A.A. and Househ, M. (2022). Machine Learning Models to Detect Anxiety and Depression through Social Media: A Scoping Review. Computer Methods and Programs in Biomedicine Update, p.100066. doi:https://doi.org/10.1016/j.cmpbup.2022.100066.
• Bezet, A. (2023). LibGuides: Research Process: Literature Gap and Future Research. [online] resources.nu.edu. Available at: https://resources.nu.edu/researchprocess/literaturegap.
• Doheny, M. (2023). Subject & Study Guides: Literature Review Guide: How to organise the review. [online] ait.libguides.com. Available at: https://ait.libguides.com/literaturereview/organise.
• Ganatra, H.A. (2025). Machine Learning in Pediatric Healthcare: Current Trends, Challenges, and Future Directions. Journal of Clinical Medicine, [online] 14(3), pp.807–807. doi:https://doi.org/10.3390/jcm14030807.
• Lai, W.-F. and Fong, M. (2024). Use of comparative research in the study of chemistry education: A systematic analysis of the literature. Heliyon, [online] 10(1), p.e22881. doi:https://doi.org/10.1016/j.heliyon.2023.e22881.
• Mahmood, S., Hasan, R., Hussain, S. and Adhikari, R. (2025). An Interpretable and Generalizable Machine Learning Model for Predicting Asthma Outcomes: Integrating AutoML and Explainable AI Techniques. World, 6(1), p.15. doi:https://doi.org/10.3390/world6010015.
• Marjan, M. (2017). A Comparative Analysis of Two Qualitative Methods: Deciding Between Grounded Theory and Phenomenology for Your Research. Vocational Training: Research And Realities, 28(1), pp.23–40. doi:https://doi.org/10.2478/vtrr-2017-0003.
• Snyder, H. (2019). Literature Review as a Research methodology: an Overview and Guidelines. Journal of Business Research, [online] 104(1), pp.333–339. doi:https://doi.org/10.1016/j.jbusres.2019.07.039.
• Sun, C., Jiang, M., Gao, L., Xin, Y. and Dong, Y. (2023). A novel study for depression detecting using audio signals based on graph neural network. Biomedical Signal Processing and Control, 88, pp.105675–105675. doi:https://doi.org/10.1016/j.bspc.2023.105675.

A. Explanation of Concept of Research Methodology

• What is Research Methodology? - At its core, research methodology is the blueprint of your research (Swarooprani, 2022). It's the systematic plan that guides how you'll answer your research question or test your hypothesis. It specifies the data needed, the techniques used to gather the information, and the procedures to analyse it. This chapter of your dissertation is all about clearly explaining that plan to your reader.
• Why is the Methodology Chapter Important? - The methodology chapter serves as a critical bridge between your research question and your findings. It's not just a formality; it's where you demonstrate the rigor and validity of your research (Dehalwar, 2024). A well-written methodology section assures your readers that your findings are trustworthy because they were obtained through a careful and systematic process. It allows other researchers to evaluate your work and, ideally, replicate your study to confirm your results. Think of it as building trust with your audience.

Research Design: Quantitative vs. Qualitative

The first and most important step in the methodology chapter is to explain your research design. Research design is the overall strategy you will use to answer your research question. There are generally two main approaches: quantitative and qualitative research (Sardana et al., 2023). Understanding the differences and choosing the right one for your study is crucial.

Quantitative Research: This type of research focuses on numerical data and statistical analysis. It aims to measure and quantify relationships between variables. Think of it as testing a hypothesis with numbers.

• Focus: Measuring and quantifying variables, testing hypotheses, and establishing relationships between them.
• Data: Numerical data, such as survey responses (using scales), experimental results, or statistical records.
• Analysis: Statistical techniques like regression, t-tests, ANOVA, and correlation.
• Goals: To identify patterns, make predictions, and generalize findings to a larger population.

Qualitative Research: This type of research explores in-depth understanding, experiences, and meanings. It emphasizes rich, descriptive data to uncover complex insights.

• Focus: Exploring experiences, perspectives, and meanings. Understanding complex phenomena in their natural context.
• Data: Non-numerical data, such as interviews, focus groups, observations, and textual analysis of documents.
• Analysis: Thematic analysis, discourse analysis, content analysis, and interpretation.
• Goals: To understand the "why" and "how" of a phenomenon, develop new theories, or gain a deeper understanding of a specific topic.

Designing which to use: Your choice between quantitative and qualitative research depends heavily on your research question and objectives. Consider what you want to learn and the kind of information you need to collect. You may also choose to combine both approaches in a mixed-methods approach.

The Research Onion Framework

Research Onion Framework | Source: ResearchGate

The Research Onion is a helpful framework for structuring and planning your research methodology. It provides a layered approach, guiding you through the key decisions you need to make about your research design (Youcef ZIDANE, 2015). Each layer represents a set of choices, starting with broad philosophical assumptions and narrowing down to the specific methods you will use. The central layer is the research question.

A breakdown of each layer is given below, moving from the outer to the inner layer

Layer 1: Research Philosophy

This outermost layer deals with your fundamental beliefs about how you view the world and how knowledge is acquired. Your philosophical stance will influence your choice of research methods. There are several key philosophies to consider:

• Positivism: This philosophy assumes that reality is objective and can be measured and observed. Research should be value-free, and the goal is to identify and test hypotheses using quantitative methods (Dudovskiy, 2019).
• Realism: Similar to positivism, realism believes in an objective reality but acknowledges that our understanding of it is imperfect. It might use both quantitative and qualitative methods.
• Interpretivism: This philosophy argues that reality is subjective and socially constructed. Research focuses on understanding the meanings and interpretations of individuals. Qualitative methods are often preferred (Pervin and Mokhtar, 2022).
• Pragmatism: This philosophy is focused on solving real-world problems and emphasizes the importance of using whatever methods are most appropriate for the research question. It is open to combining both quantitative and qualitative approaches.

Layer 2: Research Approach

This layer focuses on the process you will use to conduct your research. The two primary approaches are:

• Deductive Approach: This approach starts with a theory or hypothesis and then collects data to test it. It often involves quantitative methods. You move from the general (theory) to the specific (data).
• Inductive Approach: This approach starts with collecting data and then develops a theory or hypothesis based on the patterns observed. It often involves qualitative methods. You move from the specific (data) to the general (theory) (Sirisilla, 2023).

Layer 3: Research Strategy

This layer focuses on your overall plan and how you will collect and analyze data. Common research strategies include:

• Experiment: This involves manipulating variables in a controlled environment to determine cause-and-effect relationships (often used in quantitative research) (Lim, 2024).
• Survey: This involves collecting data from a sample of individuals using questionnaires or interviews (can be quantitative or qualitative).
• Case Study: This involves an in-depth investigation of a single case (individual, organization, event, etc.) (often used in qualitative research) (Priya, 2021).
• Action Research: This involves a cyclical process of planning, acting, observing, and reflecting, often used to solve practical problems in a specific setting (can be both qualitative and quantitative).
• Grounded Theory: This involves developing a theory based on data collected through observations, interviews, or other means (primarily qualitative).
• Ethnography: This involves immersing yourself in a culture or social group to understand their behaviors, beliefs, and values (primarily qualitative) (Naidoo, 2018).
• Archival Research: This involves analyzing existing data sources, such as documents, records, or databases (can be both qualitative and quantitative).

Layer 4: Research Choices

This layer deals with the specific choices you make about your research methods. You have three primary choices:

• Mono-method: Using a single method (e.g., a quantitative survey).
• Mixed-methods: Combining both quantitative and qualitative methods. This can involve collecting and analyzing both numerical and descriptive data (Guetterman, Fetters and Creswell, 2015).
• Multi-method: Using several methods within either quantitative or qualitative research.

Multi-method:

This layer considers the time frame of your study:

• Cross-sectional: Data is collected at a single point in time.
• Longitudinal: Data is collected over a period of time, allowing you to track changes and developments.

Layer 6: Data Collection Techniques

This innermost layer specifies the specific methods you will use to collect your data. This builds upon previous choices about methods and research strategies. Examples include:

• Surveys: questionnaires (online or paper based), interviews (structured, semi-structured, unstructured).
• Experiments: controlled experiments, field experiments.
• Observations: participant observation, non-participant observation.
• Document Analysis: analyzing documents, reports, media content, archival records.
• Focus Groups: facilitating group discussions.
• Secondary Data Analysis: using existing datasets

When and why to choose it?

When choosing the appropriate elements of the research onion, the selection should be driven by the nature of your research question and the specific objectives of your study. The research philosophy layer is important, as it dictates your viewpoint and how you approach knowledge acquisition (Saunders, Lewis and Thornhill, 2017). For example, if you believe in objective reality and the need for measurable data, you might lean towards positivism.

The choice of research approach is also crucial; if you start with a theory and aim to test it, a deductive approach is ideal. Conversely, if you're exploring a new phenomenon and want to build a theory from your observations, an inductive approach would be more suitable.

Selecting your research strategy depends on your research question and the type of data you need. For example, surveys are excellent for gathering quantitative data from a large sample, while case studies provide in-depth understanding of a specific situation (Ponto, 2015).

The decision regarding research choices, whether to use a mono-method, a mixed-methods approach, or a multi-method approach, hinges on the complexity of your research question and the diversity of perspectives you want to incorporate (Ojebode et al., 2018). Consider whether a single method will be sufficient or if you'll benefit from collecting both quantitative and qualitative data to create a comprehensive understanding.

Your time horizon, whether cross-sectional or longitudinal, will depend on the nature of the change you are investigating. Finally, the data collection techniques should align with the research strategy and your need to collect the data and analyze the data that will aid in answering your research question and meeting your research objectives.

In order to write the research methodology chapter in a proper and concise way, students should adhere to the following structure as per their research-

1. Choosing the Right Research Design:

One of the first steps in crafting your methodology is selecting an appropriate research design. This is the overall strategy you'll use to answer your research question. The most common designs include quantitative, qualitative, and mixed methods. It is critical to justify the selection of your research design by clearly linking it to your research questions and objectives.

• Quantitative: Quantitative research emphasises numerical data and statistical analysis to test hypotheses. For example, in a study like "Voice-Based Detection of Depression Using Machine Learning Techniques," a quantitative approach is used to conduct a structured evaluation of voice-based detection methods for depression. This involves experimental testing of speech datasets (DAIC-WOZ and TESS) and extracting vital voice features. These features are then analysed using Support Vector Machines (SVM) and Long Short-Term Memory (LSTM) networks, with performance evaluated using metrics like accuracy, precision, recall, and F1-score (Huang et al., 2024). Such methods align well with research questions that seek to measure and quantify relationships, such as "How can machine learning approaches make it easier to recognize depressive patterns within speech?".
• Qualitative: Qualitative research explores in-depth understanding through non numerical data such as interviews, observations, and textual analysis. It’s used to explore complex phenomena, understand perspectives, and generate new theories (Tenny, Brannan and Brannan, 2022). Qualitative methods are particularly valuable when exploring people’s emotions, behaviors, motivations and experiences, providing rich, deeper insights that quantitative data alone cannot capture. For example, a researcher might conduct in-depth interviews with healthcare providers to understand their experiences with a new electronic health record system. The goal is to uncover their perceptions of the system's usability, its impact on their workflow, and any challenges they encountered during implementation.
• Mixed Methods: Mixed methods research combines both quantitative and qualitative approaches, integrating both numerical data and rich, descriptive insights to provide a more comprehensive understanding of the research problem (Tariq and Woodman, 2017). This approach can provide a more complete picture by integrating both numerical data and rich, descriptive insights. For example, a researcher may quantitatively measure customer satisfaction scores and then conduct qualitative interviews to understand the reasons behind those scores.

Notes:

• Justify the choice of research design based on your research questions and objectives.
• Consider using the Research Onion framework to guide your methodological choices. This framework helps ensure alignment from your philosophical assumptions to your data collection techniques (Tengli, 2020). It prompts you to think about your research philosophy (e.g., positivism, interpretivism), research approach (e.g., deductive, inductive), research strategy (e.g., experiment, survey, case study), time horizon (cross-sectional or longitudinal), and data collection methods in a cohesive way

2. Research Approach:

Experimental (if Quantitative) - The experimental approach is a systematic method used to investigate causal relationships by manipulating one or more independent variables while controlling others (Podsakoff and Podsakoff, 2019). This approach allows researchers to determine the effect of specific interventions or treatments on dependent variables. A generic steps which are involved in the experimental approach are as follows.

Experimental Analysis Steps

• Data Preparation: Describe the dataset used, including how it was acquired, pre-processed, and split into training, validation, and testing sets. For example, "The DAIC-WOZ dataset will be used. The audio data will be pre-processed by removing noise, normalizing volume, and segmenting audio files into smaller, manageable chunks. The dataset will be split into 70% training, 15% validation, and 15% testing sets” (Joseph, 2022).
• Model Selection: Explain the machine learning models you'll be using. For example, "Support Vector Machines (SVM) and Long Short-Term Memory (LSTM) networks will be used due to their proven effectiveness in analysing sequential data like voice patterns."
• Model Training: Describe the training process, including the optimization algorithm, learning rate, and number of epochs. Example: "The models will be trained using the Adam optimizer with a learning rate of 0.001 for 100 epochs. Early stopping will be implemented to prevent overfitting."
• Validation and Tuning: Explain how you'll validate the model and tune its hyperparameters. For example:, the validation set will be used to tune hyperparameters such as the number of layers in the LSTM network and the kernel type in the SVM model. Grid search and cross-validation will be used to find the optimal hyperparameter settings."

Explain Data Collection Procedures: In a machine learning context, this focuses on how data is prepared and fed into the model:

• Feature Extraction: Specify the features extracted from the data. For example, "Voice features such as pitch, tone, speaking rate, and vocal intensity variation will be extracted using the OpenSMILE toolkit."
• Data Encoding: Explain how the data is encoded for input into the machine learning model. For example, "Categorical features will be one-hot encoded, and numerical features will be scaled using standardization to have zero mean and unit variance."

Outline Methods for Experimental Analysis: Describe how you'll evaluate the performance of your machine learning model:

• Evaluation Metrics: Specify the metrics you'll use to evaluate the model's performance. For example, "The performance of the models will be evaluated using accuracy, precision, recall, and F1-score on the testing set. Confusion matrices will also be generated to visualize the model's performance (Conciatori, Valletta and Segalini, 2024)."
• Statistical Significance: Explain how you'll determine if the results are statistically significant. For example, "Statistical significance will be determined using t-tests to compare the performance of the machine learning model with the baseline method, with a significance level of p < 0.05."

Data Collection (if Qualitative)

The qualitative research approach focuses on exploring in-depth understanding through non-numerical data. It is used to gain insights into complex phenomena, understand perspectives, and generate new theories (Tenny, Brannan and Brannan, 2022). This approach is particularly valuable in fields where human experiences, emotions, and motivations are central to the research question.

Data Collection Procedures: In qualitative research, various methods can be employed to gather rich, descriptive data, which are as follows-

• Surveys: While surveys are typically associated with quantitative research, they can also be designed to collect qualitative data through open-ended questions.
  • Example: A researcher might design a survey that includes both closed-ended questions (e.g., rating satisfaction on a scale of 1-5) and open-ended questions (e.g., 'What factors contribute most to your satisfaction with the healthcare system?'). The open-ended responses would be analysed qualitatively to identify common themes and insights.
• Interviews: Specify the type of interview (structured, semi-structured, or unstructured) you will conduct, as this will influence the depth of information gathered.
  • Example: In this study, semi-structured interviews will be conducted with participants who have experienced depression. The interview will include both predetermined questions (e.g., 'Can you describe your experience with depression?') and follow-up questions based on participants' responses to allow for deeper exploration of their experiences. This flexibility enables the researcher to probe for more detailed information while maintaining a focus on key topics.

Data Analysis

• Quantitative: Describe statistical techniques used (e.g., t-tests, ANOVA, regression).
• Qualitative:
  • Thematic Analysis: Explain how recurring themes were identified and coded.
  • Systematic Review: Detail the process of searching, selecting, and synthesizing existing research.
  • Other methods (e.g., content analysis).

Data Sources

• Primary Data: This refers to data you collect first-hand for the specific purpose of your study. Justify the methods used to collect this primary data (Ajayi, 2023). For example, if you are conducting interviews, explain why you chose that method over a survey. If you are conducting experiments, describe your experimental protocol in detail.
• Secondary Data: This includes existing datasets or information sources. If you're using secondary data, specify the source (e.g., government databases, published articles, other research studies) and justify its use. Explain why this data is appropriate for your research question and what limitations it might have. For example, in the “Voice-Based Detection of Depression” example, the voice recordings from datasets DAIC-WOZ and TESS serve as primary data.

Tools/Techniques/Software/Frameworks

• Specify the tools, techniques, software, or frameworks used for data collection and analysis.
• Examples:
  • Statistical software (e.g., SPSS, R) for quantitative analysis.
  • Nvivo for qualitative data coding and analysis (Wong, 2022).
  • Programming languages/libraries (e.g., Python, TensorFlow) for Machine Learning, Data Science, Cyber Security, or Networking applications.
• Explain how these tools facilitate data collection and analysis.

Ethical, Legal, Social, and Professional Considerations:

• Address ethical issues related to data collection and analysis.
• Examples:
  • Informed consent: Explain how you obtained informed consent from participants, ensuring they understood the purpose of the research, their right to withdraw, and how their data would be used
  • Data privacy and confidentiality: Describe the measures you took to protect the privacy and confidentiality of participants' data, such as anonymizing data, using secure storage methods, and limiting access to the data
  • Potential biases: Acknowledge any potential biases in your research design or data collection methods and explain how you mitigated those biases. For example, if you used a convenience sample, acknowledge that this might limit the generalizability of your findings.
  • Adherence to relevant ethical guidelines and regulations such as those from your university or professional organizations in your field. Also, address any legal and social implications of your research.
• Address legal and social implications of the research.
• Discuss professional responsibilities and standards.

References

• Ajayi, V.O. (2023). A Review on Primary Sources of Data and Secondary Sources of Data. [online] ResearchGate. Available at: https://www.researchgate.net/publication/370608670_A_Review_on_Primary_Sources_of_Data_and_Secondary_Sources_of_Data.
• Conciatori, M., Valletta, A. and Segalini, A. (2024). Improving the quality evaluation process of machine learning algorithms applied to landslide time series analysis. Computers & Geosciences, 184, p.105531. doi:https://doi.org/10.1016/j.cageo.2024.105531.
• Dehalwar, K. (2024). Basics of Research Methodology, Writing and Publication. [online] ResearchGate. doi:https://doi.org/10.5281/zenodo.11393640.
• Dudovskiy, J. (2019). Positivism - Research Methodology. [online] Business Research Methodology. Available at: https://research-methodology.net/research-philosophy/positivism/.
• Guetterman, T.C., Fetters, M.D. and Creswell, J.W. (2015). Integrating Quantitative and Qualitative Results in Health Science Mixed Methods Research through Joint Displays. The Annals of Family Medicine, [online] 13(6), pp.554–561. doi:https://doi.org/10.1370/afm.1865.
• Huang, X., Wang, F., Gao, Y., Liao, Y., Zhang, W., Zhang, L. and Xu, Z. (2024). Depression recognition using voice-based pre-training model. Scientific Reports, 14(1). doi:https://doi.org/10.1038/s41598-024-63556-0.
• Joseph, V.R. (2022). Optimal ratio for data splitting. Statistical Analysis and Data Mining: The ASA Data Science Journal, [online] 15(4), pp.531–538. doi:https://doi.org/10.1002/sam.11583.
• Lim, W.M. (2024). What is quantitative research? An overview and guidelines. Australasian Marketing Journal (AMJ), [online] 0(0). doi:https://doi.org/10.1177/14413582241264622.
• Naidoo, L. (2018). Ethnography: An Introduction to Definition and Method. An Ethnography of Global Landscapes and Corridors. [online] doi:https://doi.org/10.5772/39248.
• Ojebode, A., Ojebuyi, B.R., Oladapo, O.A. and Oyedele, O.J. (2018). Mono-Method Research Approach and Scholar–Policy Disengagement in Nigerian Communication Research. The Palgrave Handbook of Media and Communication Research in Africa, [online] pp.369–383. doi:https://doi.org/10.1007/978-3-319-70443-2_20.
• Pervin, N. and Mokhtar, M. (2022). The Interpretivist Research paradigm: A Subjective Notion of a Social Context. International Journal of Academic Research in Progressive Education and Development, 11(2), pp.419–428. doi:https://doi.org/10.6007/IJARPED/v11-i2/12938.
• Podsakoff, P.M. and Podsakoff, N.P. (2019). Experimental designs in management and leadership research: Strengths, limitations, and recommendations for improving publishability. The Leadership Quarterly, 30(1), pp.11–33.
• Ponto, J. (2015). Understanding and Evaluating Survey Research. Journal of the Advanced Practitioner in Oncology, 6(2), pp.168–171.
• Priya, A. (2021). Case study methodology of qualitative research: Key attributes and navigating the conundrums in its application. Sociological Bulletin, 70(1), pp.94–110. doi:https://doi.org/10.1177/0038022920970318.
• Sardana, N., Shekoohi, S., Cornett, E.M. and Kaye, A.D. (2023). Qualitative and quantitative research methods. [online] ScienceDirect. Available at: https://www.sciencedirect.com/science/article/abs/pii/B9780323988148000081.
• Saunders, M., Lewis, P. and Thornhill, A. (2017). Understanding Research Philosophies and Approaches. [online] ResearchGate. Available at: https://www.researchgate.net/publication/309102603_Understanding_research_philosophies_and_approaches.
• Sirisilla, S. (2023). Inductive and Deductive Reasoning | Definitions, Limits & Stages. [online] Enago Academy. Available at: https://www.enago.com/academy/inductive-and-deductive-reasoning/.
• Swarooprani, K. (2022). An Study of Research Methodology. International Journal of Scientific Research in Science, Engineering and Technology, [online] 9(3), pp.537–543. doi:https://doi.org/10.32628/ijsrset2293175.
• Tariq, S. and Woodman, J. (2017). Using Mixed Methods in Health Research. JRSM Short Reports, [online] 4(6). Available at: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3697857/.
• Tengli, M.B. (2020). Research Onion: A systematic approach to designing research methodology. [online] ResearchGate. Available at: https://www.researchgate.net/publication/357284560_RESEARCH_ONION_A_SYSTEMATIC_APPROACH_TO_DESIGNING_RESEARCH_METHODOLOGY.
• Tenny, S., Brannan, J. and Brannan, G. (2022). Qualitative study. [online] National Library of Medicine. Available at: https://www.ncbi.nlm.nih.gov/books/NBK470395/.
• Wong, L. (2022). Data Analysis in Qualitative Research: A Brief Guide to Using Nvivo. Malaysian Family Physician : the Official Journal of the Academy of Family Physicians of Malaysia, [online] 3(1), p.14. Available at: https://pmc.ncbi.nlm.nih.gov/articles/PMC4267019/.
• Youcef ZIDANE (2015). Project Change in Large Scale Engineering Projects. [online] ResearchGate. Available at: https://www.researchgate.net/publication/286453310_Project_Change_in_Large_Scale_Engineering_Projects.

This chapter details the practical steps taken to realize your project. Provide clear and concise documentation of the implementation process, ensuring reproducibility and transparency.

For Quantitative Projects: Include practical screenshots illustrating the implementation process. These screenshots should visually demonstrate key steps, configurations, or interfaces. Provide captions that explain the purpose and significance of each screenshot.

Example of Chapter 4: Implementation

Importing required libraries

Extracting and defining the features from the audio files

The code iterates through the directories of audio files organized by emotion, extracting features from each .wav file. It then appends these features and their corresponding emotion labels to two separate lists for further analysis (Tan et al., 2021).

Checking the labels present in the dataset

Label encoding the labels in the dataset. It converts the label into numeric values such as 0,1 ,2.

Checking the labels in the dataset after label encoding.

The code separates the DataFrame into two parts: features (X) and labels (y) for the machine learning model. It then splits these into training and testing sets, with 20% of the data reserved for testing, ensuring a random selection for better model evaluation (Boeschoten et al., 2023).

References

• Boeschoten, S., Catal, C., Tekinerdogan, B., Lommen, A. and Blokland, M. (2023). The automation of the development of classification models and improvement of model quality using feature engineering techniques. Expert Systems with Applications, [online] 213, p.118912. doi:https://doi.org/10.1016/j.eswa.2022.118912.
• Tan, J., Chen, Y., Liu, Z., Ren, B., Shuaiwen Leon Song, Shen, X. and Liu, X. (2021). Toward efficient interactions between Python and native libraries. arXiv (Cornell University). doi:https://doi.org/10.1145/3468264.3468541.

This chapter presents the data and provides a detailed explanation of its analysis.

• Provide a clear and concise description of the data analysis techniques employed.
• Explain the rationale behind your choice of analytical methods.
• Present the results of your analysis in a clear and organized manner, using tables, figures, and charts as appropriate.
• Interpret the findings in the context of your research questions and hypotheses. Explain the meaning and significance of your results.
• Discuss any limitations of your data or analysis.

Example of Data and Analysis Chapter

This chapter details the data used in this research and provides a comprehensive analysis of the techniques employed to detect depression through voice analysis. The study employed a quantitative research design, utilizing existing datasets to train and evaluate machine learning models (Kamiri and Mariga, 2021). The primary goal was to assess the effectiveness of voice-based analysis in identifying early indicators of depressive symptoms, comparing the performance of different machine learning techniques against established diagnostic methods.

The core data for this study comprised voice recordings obtained from established datasets specifically designed for depression research. The datasets leveraged were DAIC-WOZ (Dialogue Act and Interaction Challenge - Wizard of Oz) and TESS (Toronto Emotional Speech Set), both of which contain audio samples representing various emotional states, including depressive symptoms (Belser, 2023). These datasets were chosen for their established use in depression research and the availability of labeled data, allowing for supervised machine learning.

The methodology involved several key steps. First, the required libraries were imported to facilitate data manipulation, feature extraction, and model implementation. Second, features were extracted from the audio files (Schmidt et al., 2021). The code iterated through the audio files, which were organized by emotion type (including depressive states), and extracted relevant acoustic features from each `.wav` file. These features were then appended to lists, along with their corresponding emotion labels. These features likely included pitch, tone, speech rate, vocal intensity variation and other established acoustic features associated with depression.

Third, the labels present in the dataset were examined. To facilitate machine learning, the categorical labels representing the different emotional states were then converted into a numeric format using label encoding. For example, the different emotions could be mapped to numerical values such as 0, 1, and 2, depending on the number of emotional states captured by each dataset. The labelled data was split into features (X) and labels (y) to prepare it for model training. The dataset was subsequently divided into training and testing sets (Rácz, Bajusz and Héberger, 2021). A split of 80/20 was implemented, where 80% of the data was used for training the machine learning models, and 20% was reserved for testing. This split ensures that the models were evaluated on unseen data, providing a robust measure of their generalizability.

The choice of analytical methods was driven by the need to identify patterns in the vocal characteristics that correlate with depressive states. Support Vector Machines (SVM) and Long Short-Term Memory (LSTM) networks were selected due to their demonstrated effectiveness in classification tasks and their ability to handle sequential data, respectively. SVMs were chosen for their ability to handle high-dimensional data and their effectiveness in identifying complex patterns within the vocal features. LSTMs were selected because they excel in analysing sequential data, such as speech patterns, enabling the model to capture temporal dependencies in the voice recordings (Baig, Masud and M. Awais, 2015).

The performance of both SVM and LSTM models was evaluated using standard metrics: accuracy, precision, recall, and F1 score. The accuracy metric provides an overall measure of the model's correctness, indicating the percentage of correctly classified instances. Precision assesses the accuracy of positive predictions, while recall measures the ability of the model to identify all relevant instances. The F1 score combines precision and recall into a single metric, providing a balanced assessment of the model's performance.

References

• Baig, M., Masud, S. and M. Awais (2015). Support Vector Machine based Voice Activity Detection. pp.319–322. doi:https://doi.org/10.1109/ispacs.2006.364896.
• Belser, C. (2023). Comparison of Natural Language Processing Models for Depression Detection in Chatbot Dialogues. [online] Available at: https://dspace.mit.edu/bitstream/handle/1721.1/152710/belser-cbelser-meng-eecs-2023-thesis.pdf?sequence=1&isAllowed=y [Accessed 21 Feb. 2025].
• Kamiri, J. and Mariga, G. (2021). Research Methods in Machine Learning: A Content Analysis. International Journal of Computer and Information Technology(2279-0764), 10(2). doi:https://doi.org/10.24203/ijcit.v10i2.79.
• Rácz, A., Bajusz, D. and Héberger, K. (2021). Effect of Dataset Size and Train/Test Split Ratios in QSAR/QSPR Multiclass Classification. Molecules, 26(4), p.1111. doi:https://doi.org/10.3390/molecules26041111.
• Schmidt, L., Finnerty Mutlu, A., Elmore, R., Olorisade, B., Thomas, J., Higgins, J., Mcfarlane, E. and Kaiser, K. (2021). Data extraction methods for systematic review (semi)automation: Update of a living systematic review [version 2; peer review: 3 approved] Previously titled:Data extraction methods for systematic review (semi)automation: A living systematic review. [online] Available at: https://pmc.ncbi.nlm.nih.gov/articles/PMC8361807/pdf/f1000research-10-151999.pdf.

This chapter evaluates the outcomes of your project in relation to the research questions and existing literature.

• Clearly state the evaluation criteria used to assess the success of your project.
• Present the evidence supporting your evaluation, drawing on data and analysis from previous chapters.
• Support Your Results with Literature: Compare and contrast your findings with those reported in the literature review. Discuss how your results support, contradict, or extend existing knowledge.
• Explain any discrepancies between your findings and prior research.
• Discuss the implications of your evaluation for theory, practice, and future research.
• Acknowledge any limitations of your evaluation.

Example of Evaluation Chapter: This chapter evaluates the outcomes of the voice-based depression detection system based on two primary themes: Performance Evaluation of Machine Learning Models and Practical Implications for Mental Health Diagnostics. Each theme examines different aspects of the research findings, supporting a comprehensive understanding of the project's impact.

Theme 1: Performance Evaluation of Machine Learning Models

The evaluation of the machine learning models, Support Vector Machine (SVM) and Long Short-Term Memory (LSTM), is central to assessing the effectiveness of the voice-based depression detection system (Tao, 2024). The evaluation criteria used to assess the success of these models include accuracy, precision, recall, and F1 score.

The results indicate that the LSTM network achieved an accuracy of 85% and an F1 score of 0.82, outperforming the SVM model, which achieved 78% accuracy and an F1 score of 0.75. These findings suggest that LSTM networks are more capable of detecting depressive patterns in voice data, primarily due to their design, which enables the modelling of temporal dependencies, an essential factor for analysing speech patterns that evolve over time (Ravi et al., 2024).

This performance aligns with existing literature, supporting the notion that recurrent neural networks, particularly LSTM models, exhibit enhanced proficiency in sequential data analysis (Gan, Guo and Yang, 2024). Further to this, such results match similar studies that showcase the effectiveness of machine learning techniques in detecting depression based on vocal characteristics, reinforcing the potential for voice analysis as a reliable diagnostic tool in mental health.

Nevertheless, certain discrepancies with prior research metrics must be acknowledged. While the accuracy demonstrated in this study is commendable, it also highlights an opportunity for improvement, particularly concerning the possible overfitting of the models to the datasets used. Future research should address this by implementing cross-validation techniques and incorporating diverse datasets that better represent the population suffering from depression (Masud et al., 2025).

Theme 2: Practical Implications for Mental Health Diagnostics

The implications of this study extend to both theory and practice, particularly regarding the integration of voice analysis in clinical settings. The findings suggest that voice analysis can provide a non-intrusive, accessible method for early detection of depressive symptoms (Lin et al., 2022). This is particularly significant given the increasing demand for effective mental health interventions and the limitations of traditional diagnostic methods, which often rely on subjective assessments.

From a practical perspective, (Zafar et al., 2024) stated that the development and successful implementation of a voice-based depression detection system can facilitate earlier interventions, potentially leading to better patient outcomes. In this regard, (Zafar et al., 2024) also defined that healthcare providers may adopt voice analysis technology within existing workflows, enabling continuous monitoring of patients’ mental health status through a cost-effective means.

More to this, the current study presents a framework for other potential applications, such as the integration of additional features from multi-modal data (e.g., facial expressions or text analyses), which could enhance the robustness of depression detection systems (Boitel, Mohasseb and Haig, 2025). Future research can focus on expanding this framework, exploring how combining various data sources can create more comprehensive mental health diagnostic tools.

Despite these strengths, limitations persist. The dependency on established datasets may not fully encapsulate the diversity of vocal features associated with different depressive states (Fagherazzi et al., 2021). As well, (Hansen and Bořil, 2018) defined that the variability within collected audio samples, such as environmental noise or speaker differences, can affect the model's performance. Addressing these limitations in future research will be critical to advancing the field and ensuring the applicability of the technology across diverse populations (Hansen and Bořil, 2018).

References

• Boitel, E., Mohasseb, A. and Haig, E. (2025). MIST: Multimodal emotion recognition using DeBERTa for text, Semi-CNN for speech, ResNet-50 for facial, and 3D-CNN for motion analysis. Expert Systems with Applications, [online] 270, p.126236. doi:https://doi.org/10.1016/j.eswa.2024.126236.
• Fagherazzi, G., Fischer, A., Ismael, M. and Despotovic, V. (2021). Voice for Health: The Use of Vocal Biomarkers from Research to Clinical Practice. Digital Biomarkers, 5(1), pp.78–88. doi:https://doi.org/10.1159/000515346.
• Gan, L., Guo, Y. and Yang, T. (2024). Machine Learning for Depression Detection on Web and Social Media. International journal on semantic web and information systems/International journal on Semantic Web and information systems, 20(1), pp.1–28. doi:https://doi.org/10.4018/ijswis.342126.
• Hansen, J.H.L. and Bořil, H. (2018). On the issues of intra-speaker variability and realism in speech, speaker, and language recognition tasks. Speech Communication, 101, pp.94–108. doi:https://doi.org/10.1016/j.specom.2018.05.004.
• Lin, Y., Biman Najika Liyanage, Sun, Y., Lu, T., Zhu, Z., Liao, Y., Wang, Q., Shi, C. and Yue, W. (2022). A deep learning-based model for detecting depression in senior population. Frontiers in Psychiatry, 13. doi:https://doi.org/10.3389/fpsyt.2022.1016676.
• Masud, G.H.A., Shanto, R.I., Sakin, I. and Kabir, M.R. (2025). Effective depression detection and interpretation: Integrating machine learning, deep learning, language models, and explainable AI. Array, [online] 25, p.100375. doi:https://doi.org/10.1016/j.array.2025.100375.
• Ravi, V., Wang, J., Flint, J. and Alwan, A. (2024). Enhancing accuracy and privacy in speech-based depression detection through speaker disentanglement. Computer Speech & Language, 86, pp.101605–101605. doi:https://doi.org/10.1016/j.csl.2023.101605.
• Tao, F. (2024). Speech-based automatic depression detection via biomarkers identification and artificial intelligence approaches - Enlighten Theses. Gla.ac.uk. [online] doi:https://theses.gla.ac.uk/84055/2/2023TaoPhD.pdf.
• Zafar, F., Fakhare Alam, L., Vivas, R.R., Wang, J., Whei, S.J., Mehmood, S., Sadeghzadegan, A., Lakkimsetti, M. and Nazir, Z. (2024). The Role of Artificial Intelligence in Identifying Depression and Anxiety: A Comprehensive Literature Review. Cureus, [online] 16(3). doi:https://doi.org/10.7759/cureus.56472.

This chapter is your prospect to reflect upon your work, summarize your achievements, and evaluate your project's outcomes.

• Summarize your main findings and results, providing a concise overview of your project's key outcomes.
• Evaluate what you have achieved and how you went about it, reflecting on the methods, challenges, and successes encountered during your research.
• Conduct a critical self-evaluation of the extent to which you have achieved the things you set out to do, honestly assessing your accomplishments and shortcomings.
• Assess the extent to which you met your objectives. Were you able to answer your research question and complete the investigation?
• Acknowledge any failures to achieve everything you set out to do, especially the more advanced aspects of your project. Remember, honesty is valued over a false impression of complete success.
• Include the conclusions you have drawn from your research, providing a clear and concise summary of your key insights.
• Give recommendations based on your conclusions, suggesting practical steps or actions that could be taken based on your findings.
• Include ideas for future improvements and/or extensions of your research, identifying potential avenues for further investigation.
• Demonstrate that you have considered the commercial and economic context of your project, discussing its potential applications and implications for industry or the economy.
• Include a short section (usually one to two pages) on the management of the project, describing how you planned to allocate time at the start of the year and how it worked out in practice, reflecting on the effectiveness of your project management strategies.

Example of Conclusion Chapter

To sum up, this research successfully developed and evaluated a non-intrusive voice analysis-based system for the early detection of depression, achieving promising results with machine learning techniques. The study's main finding was that the Long Short-Term Memory (LSTM) network demonstrated superior performance compared to the Support Vector Machine (SVM) model, achieving an accuracy of 85% and an F1 score of 0.82. This outcome highlights the effectiveness of leveraging temporal dependencies in speech patterns for identifying depressive symptoms (Vandana, Marriwala and Chaudhary, 2023).

The project followed a quantitative research design, employing established datasets like DAIC-WOZ and TESS to gather and pre-process voice recordings (Ahmed et al., 2024). Key vocal features, including pitch, tone, speech rate, and vocal intensity variation, were extracted and analysed using SVM and LSTM models (Ahmed et al., 2024). The project's methods were effective in achieving its objectives, allowing the team to answer the primary research question: How can machine learning approaches make it easier to recognize depressive patterns within speech? The project met all the research objectives by gathering and pre-processing voice recordings, identifying and extracting important voice features, implementing machine learning models, and comparing them.

A comprehensive literature review identified research gaps and guided the study's methodology, which included an extensive review of existing literature, experimental testing on speech datasets, and a comparison of the voice analysis method with standard practice. Nevertheless, a limitation was the dependence on existing datasets, which, while valuable, may not fully represent the diversity of depressive states. In addition, the quality of the audio recordings could have introduced variability (Mundt et al., 2025). Despite these limitations, the study provided insights into how vocal characteristics can serve as reliable biomarkers for depression detection and monitoring treatment responses. The core conclusion is that voice analysis can be a valuable tool for depression detection, offering an objective assessment that complements traditional diagnostic methods, especially when LSTM networks are employed.

Based on these conclusions, it is recommended that further research focuses on larger and more diverse datasets and explores the integration of multi-modal data to improve the generalizability of the models. Future studies should incorporate more advanced feature extraction techniques and investigate the potential of voice analysis with other diagnostic tools (Alper Idrisoglu et al., 2023). This project has commercial and economic implications, as the development of a voice analysis system could enhance diagnostic accuracy and facilitate early intervention, offering a low-cost and scalable solution for mental health screening.

Project management was important for success which was fulfilled with Microsoft Project Software (Tea Borozan, Golub Marković and Petar Stanimirović, 2023). The initial plan involved allocating time for literature review, data acquisition, model training, and evaluation. Time allocation was followed well and allowed the project to stay on schedule. Future improvements could involve more diverse datasets and the inclusion of multi-modal data (e.g., facial expressions and text analysis).

References

• Ahmed, A., Alam Khondkar, M.J., Herrick, A., Schuckers, S. and Imtiaz, M.H. (2024). Descriptor: Voice Pre-Processing and Quality Assessment Dataset (VPQAD). IEEE Data Descriptions, pp.1–8. doi:https://doi.org/10.1109/ieeedata.2024.3493798.
• Alper Idrisoglu, Ana Luiza Dallora, Anderberg, P. and Berglund, J. (2023). Applied Machine Learning Techniques to Diagnose Voice-Affecting Conditions and Disorders: Systematic Literature Review. Journal of Medical Internet Research, [online] 25, pp.e46105–e46105. doi:https://doi.org/10.2196/46105.
• Mundt, J.C., Snyder, P.J., Cannizzaro, M.S., Chappie, K. and Geralts, D.S. (2025). Voice acoustic measures of depression severity and treatment response collected via interactive voice response (IVR) technology. Journal of Neurolinguistics, 20(1), pp.50–64.
• Tea Borozan, Golub Marković and Petar Stanimirović (2023). Navigating Project Success - A Deep Dive Into The Influence Of MS Project. XIV Conference of Business and Science SPIN’23 Digital and Green Economy Development. [online] Available at: https://www.researchgate.net/publication/377746945_Navigating_Project_Success_-_A_Deep_Dive_Into_The_Influence_Of_MS_Project.
• Vandana, Marriwala, N. and Chaudhary, D. (2023). A hybrid model for depression detection using deep learning. Measurement: Sensors, 25, p.100587. doi:https://doi.org/10.1016/j.measen.2022.100587.