What is the right research methodology for an AI postgraduate thesis?
The right research methodology AI postgraduate students should choose depends on the research question, available data, and time horizon. Systematic literature review (SLR) suits maturity-mapping and gap analysis, while experimental research suits novel model or dataset contributions. Most strong Malaysian theses combine an SLR with at least one experimental study.
Dr. Muhamad Hariz Muhamad Adnan, Senior Lecturer at Universiti Pendidikan Sultan Idris (UPSI), supervises Master and PhD candidates in explainable AI, AI in education, and precision agriculture, and guides students through this choice early in candidature.
When should a Malaysian postgraduate choose a systematic literature review?
A Malaysian postgraduate should choose a systematic literature review when the field is fragmented, evidence is scattered across venues, or stakeholders need a clear map of what has been tried. SLR works well for emerging topics like explainable AI in education, where synthesis itself is the novel contribution Malaysian universities recognise.
Strong SLR signals
- Question begins with “what is known about…” or “which methods have been used for…”
- Limited or restricted access to primary data
- Need for a high-quality publication early in candidature
- Field is large enough to have studies but lacks synthesis
- Student is strong in writing and weaker in coding
When should a Malaysian postgraduate choose experimental research?
A Malaysian postgraduate should choose experimental research when the gap is a specific model, dataset, or technique that can be tested against a baseline. Experimental work suits AI students who can code, access data, and want to publish in top venues such as IEEE Access, Scopus-indexed journals, or AI/ML conferences.
Strong experimental signals
- Question begins with “does X improve Y” or “how well does X perform”
- Access to labelled data or willingness to collect it
- Comfort with Python, PyTorch, TensorFlow, or scikit-learn
- Clear baseline models to compare against
- Need for original technical contribution for PhD
How do the two methodologies compare practically?
SLR and experimental methodologies differ in data, tooling, output, and risk. SLR depends on databases like Scopus, Web of Science, and IEEE Xplore, while experimental work depends on datasets, GPUs, and code. Each has different risks: SLR risks selection bias, while experimental risks data leakage and reproducibility issues.
| Dimension | Systematic Literature Review | Experimental Research |
|---|---|---|
| Primary input | Published studies | Datasets and code |
| Tools | Scopus, Web of Science, NVivo, Rayyan | Python, PyTorch, GPUs, MLflow |
| Typical duration | 6–12 months | 9–18 months |
| Common output | Q1/Q2 review article | Original model + journal/conference paper |
| Main risks | Selection bias, low rigour | Data leakage, weak baselines |
| Best for | Master, early PhD | PhD, applied Master |
How should the two be combined in a Malaysian PhD?
A strong Malaysian PhD typically uses an SLR in the first year to define the gap, followed by one or two experimental studies in years two and three. This sequence yields at least two journal papers, satisfies UPSI and other Malaysian university viva expectations, and reduces the risk of choosing a saturated topic.
Recommended three-paper PhD sequence
- Year 1: SLR identifying gaps and consolidating evidence
- Year 2: Experimental study 1 – baseline and proposed method
- Year 3: Experimental study 2 – extension, real-world validation, or explainability
- Thesis chapters mirror the three published papers
What rigour standards should Malaysian postgraduates meet?
Malaysian postgraduates should follow PRISMA 2020 for SLRs and clear baselines, ablation, and statistical testing for experiments. UPSI examiners increasingly expect open code, open data where possible, ethics review for human subjects, and PDPA-aligned data handling, regardless of methodology chosen.
- SLR: PRISMA 2020 flow diagram and protocol pre-registration
- Experimental: stratified train/validation/test split with seed control
- Statistical tests with effect sizes, not just p-values
- Open source code on GitHub where IP allows
- Ethics approval for any human data
- PDPA-aligned consent and data minimisation
How does Dr. Muhamad Hariz supervise methodology choice at UPSI?
At UPSI, Dr. Muhamad Hariz holds a methodology clinic in the first month of candidature, where each student maps their research question against feasibility, novelty, and time. Students leave with a draft proposal, a chosen methodology, and a publication target appropriate for Master or PhD level under Malaysian university expectations.
Frequently Asked Questions
Is an SLR enough for a PhD in Malaysia?
An SLR alone is usually not enough for a PhD in Malaysia, though it works for Master degrees. Most Malaysian universities, including UPSI, expect doctoral candidates to make an original empirical or theoretical contribution beyond synthesis. A combined SLR plus experimental design is the safest, most publishable path.
Can I publish a Scopus paper from an SLR?
Yes, you can publish a Scopus-indexed paper from an SLR if you follow PRISMA, address a clear gap, and contribute a novel framework or taxonomy. Q1 and Q2 journals in AI in education, explainable AI, and applied machine learning regularly accept rigorous reviews from Malaysian postgraduates.
Do I need a GPU for an experimental AI thesis?
You usually need GPU access for an experimental AI thesis, but you do not need to buy one. UPSI and most Malaysian public universities offer shared GPU resources, while Google Colab Pro and Kaggle provide affordable cloud options. Dr. Muhamad Hariz helps students plan compute budgets in their proposals.
How long should an AI Master thesis at UPSI take?
An AI Master thesis at UPSI typically takes 18 to 24 months full-time, depending on whether it is research mode or coursework with project. Strong students complete an SLR plus one experimental study and publish at least one Scopus-indexed paper before submission under supervision.
How do I apply for AI postgraduate supervision at UPSI?
To apply for AI postgraduate supervision at UPSI, prepare a one-page research idea and contact the Faculty of Computing and Meta-Technology directly. Visit drhariz.com to enquire about supervision under Dr. Muhamad Hariz, or read more on the blog for sample proposals.
Dr. Muhamad Hariz Muhamad Adnan is a Senior Lecturer and Acting Deputy Dean at Universiti Pendidikan Sultan Idris (UPSI), HRD Corp Certified AI Trainer, and digital transformation consultant. For AI training or postgraduate supervision enquiries, visit drhariz.com or read more on his blog.
