Guidence for Research/Project Students on Research Steps

This page is designed to support research and project students, primarily those in the 3rd or 4th year of their BSc, URSS, or enrolled in MSc taught or research programmes. The projects I supervise are typically in the field of computer science, with a focus on data analytics and recommender systems. While each topic is unique, the projects generally follow similar research steps throughout their lifecycle.

The purpose of this page is to provide an overview of these key stages. More in-depth discussions tailored to your specific research topic will take place during our meetings.

To book a meeting with me, please go to this pageLink opens in a new window.

Coming up with an idea

Identifying a compelling idea to address a research or social problem and structuring it into a coherent project is often the most challenging part of the research process. Once the foundation is solid, implementation typically becomes a more straightforward task.

Reflect on your interests, skills, and career goals.
Look into emerging technologies or real-world problems.
Explore previously completed projects for inspiration.

BEAR IN MIND: Your project has to be unique, answering novel questions, and novel application of methodology.

Project Specification: Narrowing down your idea as a research topic

Many students think finding the topic and saying what you will have as an output naturally brings the technical steps too. However, when you start thinking about what, why and how, you may feel stuck. Identifying the details of the project takes long time and this phase is essential for a successful implementation. Do not underestimate the cognitive effort and time needed in this phase, allow sufficient time.

First step: Literature Review

For the literature review, you need to understand that we investigate what others are doing, what is missing, and what are the potential research gaps that we can address to contribute to science. So that we can phrase our problem statement, objective of the research as well as research questions.

To facilitate this process, I strongly recommend that you prepare a literature review in table format.

TODO: The literature review should not just summarise papers; it should critically analyse the literature. A helpful approach is to create a comparison table that includes at least 10-15 academic sources relevant to your topic. This will clarify existing literature, identify gaps, and highlight your contributions. You can use www.googlescholar.com or www.semanticsholar.org.

Your table should include columns such as: Name, Objective, Target, Method/Model/Tools Used, Arguments/Features, Data Source, Performance, Advantages, Disadvantages, Comments, etc. Feel free to adjust the columns based on your research focus. For examples, refer to Tables 4 and 5 in my work (Sunar et al.) from my PhD. You can also check Table 1 in Ivanescu et al. to see how to highlight the novelty of your method.

Once you have a complete table, it will greatly assist in writing your thesis and help clarify your methodology.

Second step: Stating your problem, goal, objectives, research questions

You need to have a very clear on sentence "research problem statement". Based on this problem, you should identify 3-5 SMART (specific, measurable, achievable, relevant and time-bound) objectives which form your research questions. These will define how to measure your success when you complete the project.

Third step: Methodology

You analysed the literature by this step and now it is time to synthesise the information you've got so far and make informed decision about your methodology based on your defined objectives and research questions.

We expect that every research project should propose a novel idea . This does not mean that you have to invent a big thing which revolutionise the world. The science is cumulative which means we will use what other people have done so for to improve and adapt. Let's say you can adapt an algorithmic idea which is used in another context to your context. You can modify a proposed solution to your data and customers' needs. You can merge different ideas to propose a novel use etc.

At this stage you may not have a very detailed technical plan but I expect a general idea and a high level methodological diagram showing how you system flowchart looks like.

Data Selection

This is a very essential step as it may also change your methodological choices. You need to find a proper dataset with correct license which allows you to use for academic research purposes. You may like to collect data by your own, in this case, we may need to apply for ethical approval. Please note, not all websites allow you to scrap data from their websites. You may need to contact them for approval. Otherwise, it will be unethical and we cannot use any data unethically collected no mater how good the data is.

System Diagram

You each have great ideas, a general understanding of your data and goals but most of this remains in your head or written down in paragraphs. When it comes to implementation, things can become unclear or won’t work as expected. That’s where system diagrams become essential.

Think of your block/system diagrams the bridge between your idea and the machine: it explains how your system will process data and function technically. It also communicates your approach clearly to others. You will definitely use these diagrams in your presentation and reports.

Below are some useful examples to guide you:

Publication from an MSc Thesis (our department)Link opens in a new window:

See Figures 1, 2, and 3 and how they explained them in the text.
RecommenderSystems Overview (NVIDIA)Link opens in a new window:

This is not a specific proposed model, general structure of RS
ArchitectureLink opens in a new window Example (Alibaba):
Prediction Model (IEEE Access)Link opens in a new window:

They simplified the look.

These may not be exact matches to your projects, but observe how inputs, components, and outputs are mapped visually. Your contribution may not be a full system; it could be a single novel element like an embedding model but explain how it fits in. Please do look into papers, how they proposed their systems, especially the ones using the same algorithms with you. If you are going to use number of features together, think about how you will merge them, is there a better way to do it?

Forth step: Evaluation

Quantitative evaluation (Numeric metrics): You need to identify what other research uses as a metric to evaluate similar kind of research. You need to make sure that you can mathematically measure and evaluate your results to assess your achievements towards your objectives.

Qualitative evaluation (e.g. human evaluation): Your research may require asking people for their opinion on your product. In this case, we will recruit people (ethical approval needed) to ask. This evaluation only cannot be used as an evidence for success of your system, proposed method, and technical contribution. Both numeric and human evaluation can be used together.

Fifth step: Time table

Break down your methodology into steps (work packages). Some parts will require longer time and some parts will be more challenging to implement. You need to realistically design your timetable thinking holidays, coursework submission, required labour for your work packages. Students are usually very optimistic about how they will use time - so allow your plan at least 2 weeks buffer!