RESEARCH METHODOLOGY

Objectives of Research
 Research is a systematic and scientific process of inquiry aimed at generating new
knowledge, validating existing knowledge, and solving problems.
 Objectives of Research:
 • To identify gaps in knowledge and propose solutions.
 • To generate new theories and models.
 • To validate or refine existing theories.
 • To improve practices, policies, and decision-making.
 • To drive innovation and technological advancement.
 • To develop problem-solving and critical thinking skills in researchers.

Types of Research
1. Basic Research (Fundamental Research)

 Conducted to increase knowledge and understand underlying principles without

immediate practical application.

 Theory building, concept development, discovering new facts.

 Example

o Studying the mathematical properties of new cryptographic algorithms,

without immediate concern for deployment.

o Research on the behavior of quantum bits (qubits) to understand how

quantum computers might process data.

2. Applied Research

 Focused on solving specific, practical problems using existing theories or

knowledge.

 Direct application in real-world problems.

 Example
 o Developing an AI-based diagnostic tool for detecting lung cancer using
CT scans.

o Creating a cybersecurity intrusion detection system based on machine

learning models.

3. Exploratory Research

 Conducted in areas where little is known, aiming to explore new ideas and
directions.

 Gain insights, generate hypotheses, set groundwork for future research.

 Example

o Exploring the use of brain-computer interfaces (BCI) to control devices

using neural signals.

o Investigating how people interact with metaverse platforms when little

structured research exists.

4. Descriptive Research

 Provides systematic and factual descriptions of phenomena without testing cause-

effect.

 Answer “what is happening” rather than “why it is happening.”

 Example

o Conducting a survey of online learning platforms to describe usage

patterns among students.

o Analyzing social media datasets to describe trends in misinformation

spreading.
5. Experimental Research

 Involves manipulating variables in controlled conditions to establish cause-effect

relationships.

 Testing hypotheses by comparing results across controlled and experimental

groups.

 Example

o Running experiments where one group of users uses a new

recommendation algorithm while another uses the old one, then comparing
engagement.

o Testing the effect of a new compression algorithm on file transfer speeds

under different network conditions.

Relevance of Research in Computer Science

 Advances emerging technologies such as AI, ML, IoT, and quantum computing.
 Improves system performance through efficient algorithms and architectures.
 Addresses pressing issues like cybersecurity, privacy, and ethics.
 Enables innovation in healthcare, robotics, education, and automation.
 Supports data-driven decision-making and industry 4.0 transformations.

Research Paradigms
 Quantitative: Uses numerical data, statistical tests, objective conclusions.
 Qualitative: Focuses on understanding meaning, experiences, and context.
 Mixed Methods: Integrates both quantitative and qualitative for comprehensive
insights.

Quantitative Qualitative
Aspect Mixed Methods
Research Research
Non-numerical
Nature of Numerical, Both numerical +
(text, audio,
Data measurable narrative
images)
Deductive (test Inductive Combination (test +
Approach
theory) (generate theory) explore)
Measure, test, Understand Provide holistic
Goal
predict meaning, context understanding
 Quantitative Qualitative
Aspect Mixed Methods
Research Research
Surveys, Interviews, case Surveys + Interviews,
Methods experiments, studies, focus Experiments +
statistics groups Observations
Sample
Large Small Both (large + small)
Size
Statistical, Thematic, Integrated (stats +
Analysis
mathematical narrative themes)
Generalizable, Rich, detailed Balanced, validated
Outcome
objective insights findings
Studying chatbot
ML model AI ethics
Example performance (logs +
accuracy testing interviews
feedback)

Scientific Method and Research Process

 Steps:
 1. Observation.
 2. Problem Identification.
 3. Literature Review.
 4. Hypothesis Formulation.
 5. Research Design (Qualitative/Quantitative/Mixed).
 6. Data Collection.
 7. Data Analysis.
 8. Interpretation of Results.
 9. Conclusion and Reporting.

Research Problem and Hypothesis

 Research Problem: An issue, knowledge gap, or contradiction requiring study.
 Hypothesis: A tentative statement predicting the relationship between variables.
Problem Identification and Problem Statement
 Problem Identification: Recognizing a significant issue through observation,
literature, or practice.
 Problem Statement: A clear, concise, and researchable description of the problem.

Problem Identification vs. Problem Statement

Problem Identification
The process of recognizing that an issue exists which needs to be studied.
Sources: Observation, literature review, existing gaps, or professional practice.
Purpose: To spot a broad area of concern worth investigating.
Example
While reviewing literature, a researcher notices that existing deepfake detection
models struggle when audio + video are combined.
This recognition → Problem Identification: “Current deepfake detection methods
are not effective in multimodal scenarios.”
Problem Statement
A clear, concise, and researchable description of the identified problem.
Purpose: Defines the scope and direction of the research, turning a broad issue
into a specific, actionable research problem.
Example
From the above identification →
Problem Statement:
“Despite significant progress in deepfake detection, most existing methods focus
on unimodal data (either video or audio). These methods show reduced accuracy
when detecting deepfakes that combine multiple modalities. There is a lack of
robust multimodal approaches that integrate visual, audio, and textual features for
reliable detection.”

Framing Research Objectives and Questions

 Objectives must be SMART: Specific, Measurable, Achievable, Relevant, Time-
bound.
 Research Questions: Precise, focused, answerable, and aligned with objectives.

Research Objectives
 Broad statements describing what the researcher aims to achieve.
Must be SMART:
o S → Specific – Clearly defined, not vague.
M → Measurable – Can be evaluated with data.
A → Achievable – Realistic within available resources.
R → Relevant – Aligned with the field and problem.
T → Time-bound – Feasible within a defined timeline.

Example
General Aim: Improve accuracy of multimodal deepfake detection.
SMART Objectives:
1. To analyze existing unimodal and multimodal deepfake detection
approaches (Specific).
2. To design and implement a transformer-based hybrid model combining
audio, visual, and text modalities (Measurable & Achievable).
3. To evaluate the model’s performance against benchmark datasets within
12 months (Relevant & Time-bound).

2. Research Questions

Narrow, precise, and answerable questions derived from objectives.

Purpose: Guide the research process and ensure alignment with objectives.

Examples (linked to above objectives):

 RQ1: What are the major limitations of unimodal deepfake detection models?
 RQ2: How can transformer-based architectures improve multimodal deepfake
detection?
 RQ3: Does integrating audio, visual, and text features increase detection
accuracy compared to unimodal models?

Formulation of Hypotheses – Types

 Null Hypothesis (H0): No relationship/effect.
 Alternative Hypothesis (H1): There is a relationship/effect.

Aspect Null Hypothesis (H₀) Alternative Hypothesis (H₁)

There is an effect, difference, or
Meaning No effect, no difference, no relationship
relationship
Role Default assumption (to be tested) Opposes H₀ (researcher’s
 Aspect Null Hypothesis (H₀) Alternative Hypothesis (H₁)
claim)
Statistical
Researchers try to reject H₀ Accepted if H₀ is rejected
Test
No accuracy difference between unimodal Multimodal deepfake models
Example
& multimodal deepfake models have higher accuracy
A new drug has no effect on blood The new drug reduces blood
Example
pressure pressure significantly

Criteria of a Good Hypothesis

 Should be clear and concise.
 Must be testable and falsifiable.
 Consistent with existing theories.
 Predicts expected outcomes.

Hypothesis Testing – Logic, Errors, and Significance

 Logic: Compare observed data with hypothesis expectations.
 Errors:
 • Type I Error: Rejecting true null hypothesis.
 • Type II Error: Accepting false null hypothesis.
 Significance: Set threshold (p-value) for decision-making.

Research Planning and Flowcharting

 Scope: Define boundaries and coverage.
 Limitations: Acknowledge constraints in time, data, or resources.
 Timeline: Plan milestones and deadlines for stages.
 Flowcharts: Used for representing steps in research and algorithms.
Tools and Best Practices for Visual Representations
 Tools: Lucidchart, Draw.io, Canva, MS Visio.
 Best Practices:
 • Use simple and consistent shapes.
 • Maintain clarity with readable text.
 • Apply consistent color schemes.
 • Avoid clutter and overlapping elements.
 • Align diagrams with research steps logically.