Document Template (make sure you read it!

MSc Research Project

Data Analytics

Forename Surname
Student ID: XXX

School of Computing
National College of Ireland

Supervisor: Dr Giovani Estrada

 National College of Ireland
Project Submission Sheet
School of Computing

Student Name: Forename Surname

Student ID: XXX
Programme: Data Analytics
Year: 2022
Module: MSc Research Project
Supervisor: Dr Giovani Estrada
Submission Due Date: 20/12/2022
Project Title: Document Template (make sure you read it!)
Word Count: XXX
Page Count: 11

I hereby certify that the information contained in this (my submission) is information
pertaining to research I conducted for this project. All information other than my own
contribution will be fully referenced and listed in the relevant bibliography section at the
rear of the project.
ALL internet material must be referenced in the bibliography section. Students are
required to use the Referencing Standard specified in the report template. To use other
author’s written or electronic work is illegal (plagiarism) and may result in disciplinary
action.

Signature:

Date: 19th March 2024

PLEASE READ THE FOLLOWING INSTRUCTIONS AND CHECKLIST:

Attach a completed copy of this sheet to each project (including multiple copies). □
Attach a Moodle submission receipt of the online project submission, to □
each project (including multiple copies).
You must ensure that you retain a HARD COPY of the project, both for □
your own reference and in case a project is lost or mislaid. It is not sufficient to keep
a copy on computer.

Assignments that are submitted to the Programme Coordinator office must be placed
into the assignment box located outside the office.

Office Use Only

Signature:

Date:
Penalty Applied (if applicable):
 Document Template (make sure you read it!)
Forename Surname
XXX

Abstract
Abstracts are useful for indexing and helping people decide whether or not to
read the entire report1 . Expected structure: 1. show the big picture; 2. problem
description; 3. proposed solution; 4. main results; 5. implications and/or who can
benefit from these results. In short, do not get lost in details or vague claims.

Note: overview
The present annotated research template will help learners to prepare their submissions.
It is essentially the official template shared via Moodle with a few added comments. It
has a number of steps to remind you about the expected content and location for each of
elements of a research report. You have to make abundantly clear the following content:

✓ What is your research about? (from big picture to research gap to research question)

✓ Why is that important? (significance of selected research question; why is research

worth doing?; value to the world)

✓ What will the scope and limitations of your research be?

✓ How do you plan to address the research question? (logical steps to address your
research question; your research goals)

The number of steps are meant to be minimalist rather than an extensive checklist.
A few common pitfalls and difficulties are also mentioned along the way.

1 Introduction (the what and why of your research)

Your 20-page manuscript should have the logical flow of an academic report. For the
Introduction, set the context and present the big picture. Start big, then narrow down
your research area.
You have heard the expressions “take a step back” and “look at the big picture” or
“bird’s eye view”. In essence, talk about your area of interest without drowning readers in
too much details. Go to the point. One or two drawings should help you to contextualise
the field of interest and then narrow down your research field, for instance as shown in
Figure 1a.
1
Write it at the end of your work, when you know what your main results are. Example https:
//www.nature.com/documents/nature-summary-paragraph.pdf

1
Hints
✓ Draw some nice diagrams to show what your idea is about. Put some boxes together
to show how your proposed solution look like, as seen in Figure 1b.

✘ Screenshots appear unprofessional in academic documents. Think twice before in-

serting that horrible screenshot. There is no reason why tables are included as
screenshots, you can type them! Learn how to create LATEX tables2 .

(a) Set the context (b) Typical workflow

Figure 1: Include some meaningful captions: (a) what should we look at?; (b) what should
we look at? Remember to reference every single figure in your document!

Note: Organise text for reading

Please avoid unnecessary abbreviations and acronyms – and, if you use them, write down
their meaning. Unnecessary jargon will only obscure the main content and put readers
off your manuscript. It is best to keep definitions to a bare minimum.
It is a good idea to read about academic writing. Each discipline has writing conven-
tions, basic abbreviations and acronyms that you and the reader expect to be familiar
with. In a nutshell, avoid writing wordy, dense sentences. Express only one idea in each
sentence, which you can then piece together to create short paragraphs. Cover one topic
per paragraph.

1.1 Research Gap (things that can be improved)

Here you can talk about how interesting your area is, and then make the reader be aware
of typical problems, trade-offs and issues. Here, is where you see an opportunity to make
things faster, cheaper, less labour intensive, more cost effective, more time efficient, and
so on. Use a couple of paragraphs to slowly move from the big picture and the research
gap.
2
See for instance https://www.tablesgenerator.com/

2
 Leave the tons of references for Section 2. Here, just bring the attention to opportunity
areas, and cite a couple of recent surveys pointing to how active this area is. In the next
section you will narrow down the area of interest to just one single topic. You are
justifying the research: the what and why something has to be studied.

1.2 Research question (the core of your document)

The research question is a very important part of the document. The entire document
gravitate towards the research question. The research question should give you enough
room to showcase your knowledge and skills.
Some research questions originate from papers, Kaggle submissions, or just browsing
through the Internet. It is fine. You know the context, probably have working code and
dataset – but it is not clear what you can do research about. Unknowns will normally
arise once you go deeper into the area of interest.

Hints
✘ Questions like “can I apply image processing techniques to my image dataset?” are
too simplistic. The answer is a trivial “yes”. Why do you want to write 20 more
pages about it?

✓ Look for little known aspects of the process you got familiar with. Challenge those
“default” values or widely accepted assumptions. It is time to explore alternatives!

The point is to target just one single aspect. That way, you should be able to narrow
down a question to something that can potentially generate knowledge in a quantifiable
and verifiable manner. Table 1 shows examples of research questions.
Valid research question(s) will not exist in a vacuum. So, please write down any
tentative answers (i.e. hypotheses) that you want to validate, as well as any working
assumptions. Depending on the question(s) you could remark that you will work with
multi-class problems (or binary classification), focused on certain country or region, time
period, etc. By doing this, you warn the reader that results may hardly be generalisable
due to sample size, data collection, technique of choice, etc.

1.3 Research objectives

Here you should briefly write the steps you are about to take to answer the research
question(s). It can be a simple list of most important steps and/or a flowchart showing the
concrete steps you will take to achieve the goals. Valid entries here would be, depending
on your project, data collection, pre-processing, augumentation, generation of baseline
classifier, etc. Please, just write your most difficult steps, the ones you spent weeks of
your time (your milestones).

1.4 Outline
Finally, you can now close Section 1 by outlining the structure of the report, for instance:
The remainder of the report is organised as follows. Section 2 presents the relevant theory
and works closely related to the proposed one. Section 3 describes details of the proposed
approach. In Section 4, we describe the techniques that are used to address the problem,

3
Table 1: Mock-up research questions. The question should focus on a systematic study
of a topic, beyond the trivial “yes” or “not” answers. Notice that table captions appear
above the table!

What? How?
What effect do pooling layers have on How do ResNET and DenseNET com-
the classification accuracy of VGG123? pare in training time and model com-
plexity for image classification?
To what extent can activation functions How have CNNs impacted the devel-
be tuned to improve autoencoders? opment of time series over the past 10
years?
What effect does stacking layers have How can concatenate layers be effect-
on accuracy and training time of VGG ively used to improve the classification
networks? accuracy of CNNs?
What impact have batch normalisation How have CNN-based time series and
layers on the convergence of validation ARIMA algorithms compare in terms
loss values in Inception networks? of accuracy and computing time?
What are the most important factors How do you combine data augumenta-
driving the classification of stars? tion steps to improve the efficiency of
autoencoders?
What are the challenges and limita- How do you select the best strategy to
tions of GANs in the generation of syn- forecast non-seasonal time series?
thetic tabular data?
What are the most effective modules in How might data augmentation impact
Inception networks and what can be re- the overall classification accuracy of
used for other networks? ResNet networks?

as well as all proposed test cases. The two novel methods are presented in Section 5, while
evaluation results appear in Section 6. Finally, we provide conclusions and discussions of
future research directions in Section 7.

2 Related Work (critical review of closely related pa-

pers)
Now situate your work in the academic literature; this entails a critical (positive, negative,
helpful) review of closely related work. What work? Work that is relevant to the research
question(s). If you can not find similar work, you have not looked hard enough. Ideally,
you want to be reading around 50 papers; of which at least 25 should appear in the
thesis itself. Note that URLs are not references, they are footnotes.3

Hints
✘ Do not copy/paste sections from your research proposal (RIC CA2) blindly. It is
unlikely the content is suitable at this stage.
3
Like this one: http://www.ncirl.ie

4
 ✓ Only include references aligned to your research question.

You are expected to provide a critical/analytic synthesis of existing research. Com-

ment on the strength and weakness/limitation of work in reviewed papers.

Hints
✘ Do not start a sentence with a citation. The citation should come at the end of the
sentence Feng and Buyya (2016). Or, the citation should appear Feng and Buyya
(2016) in the middle of a sentence.

✓ The content sections of your report should of course be structured into subsections.

2.1 Subsection 1
If you decide to use Microsoft Word, please pay attention to the correct citation style.
Also, the References section, at the end of the document, is not numbered. A common
mistake is to write incomplete references, place them in the beginning of the sentences or
write trivial sentences to use them Beloglazov and Buyya (2015); Gomes et al. (2015).

LATEXnote
• It is worth mentioning that your citations must be included in the refs.bib file. It
is in bibTEXformat, open it up to see how it looks like and populate it with your
own references.

• Change your name, studentID, and document title in the file researchProject.tex,
see for instance placeholders “Forename” and “Surname”.

• A nice little way of leaving yourself notes and reminders: (Use todo to for doc- ToDo
ument reminders)

2.2 Subsection 2
Use figures to make your content more appealing, but always cite the source of your
figures. Make sure you always describe what the figure is about in your document and
figure captions, see for instance Figure 2.

Figure 2: Notice figure captions appear below the figure. It should summarise a result
or major finding – guide the reader to that interesting bit, without having to read the
entire report! Cite the source, if appropriate.

After writing the previous subsections, it is now time to summarise your main ref-
erences into a table. In Table 2 an example is provided. Please type the table, do not

5
Table 2: Notice table captions appear above the table. It should summarise a result or
major finding – What makes your work different to published papers?

Paper Description & Drawbacks Accuracy %

Beloglazov and Buyya (2015) proposed X;
used very limited data 93.65
Feng and Buyya (2016) introduced Y;
did not show accuracy on validation set 90.01
Gomes et al. (2015) said to improve Z;
did not provide details to validate claims 92.50
Kune et al. (2016) proposed a novel W;
only tested binary classifiers 98.99

insert screenshots. It is a recap of what you have said, so focus on the very state of the
art items.
Provide a brief, crucial review to contrast with your own work. Say what you work is
about, and what is not. For example, from Table 2 the reader will understand that your
research:

• will use large scale data;

• will use a validation set;
• will provide enough implementation details; and
• will focus on multi-class problems.

Finally, the literature review should end up with a paragraph saying why previous
solutions are not adequate and thus justifies the need for further research (yours!). Typ-
ically, something along the lines of: “X thing has never been discussed previously”, “X
thing is little known”, “nobody knows how to do X”, etc. Your justifications have to be
aligned to your research question(s) and objectives (Sections 1.2 and 1.3).

2.3 References
The correct style for your references is the one from this template. In other words, do
not change the template. If you are going to use Word, please follow the style closely.

✓ References in tables, e.g. Table 2, look good using the command “cite”. The result
appears as: Beloglazov and Buyya (2015).
✓ If you insert the citation within a paragraph, the command is “citep”. The result
will appear as: (Beloglazov and Buyya; 2015).
✓ Let the reader know where you got the information from. In other words, all claims
must be backed up with a proper reference.
✓ The bulk of your references should come from reputable sources (IEEE, Science-
Direct, etc) and ideally from the last five years.
✘ Do NOT start your sentence with a citation. The citation should be placed in the
middle or end of your sentence.

6
3 Methodology (either KDD or CRISP-DM)
You will of course want to discuss your research as well as evaluation methodology.
Otherwise, how will your examiners know that you have followed a scientific, systematic
process and rationalised your choice of evaluation. Note that it may also be useful to
base decisions in this section off your discussion of related work in Section 2.

Hints
✘ Too much detail here obscures the focus of your document. Implementation details
are placed elsewhere, either in the Configuration Manual or in Section 5.

✓ Draw a diagram and enumerate the boxes. Then, explain every single box, from
raw data to pre-processing to main results.

You need to give a completely accurate description of the research procedure you
followed, equipment used, the technique(s), applied, set-up of scenarios/case studies run,
etc. Adapt either KDD4 or CRISP-DM5 , Figure 3, to your own needs. Hints: KDD tends
to be more linear, “one-off” effort to identify or understand a pattern. CRISP-DM is a
more general methodology that includes business and data understanding steps. If the
research work includes substantial cloud deployment efforts it should follow CRISP-DM.

(a) KDD (b) CRISP-DM

Figure 3: Two reference methodologies: (a) Knowledge Discovery in Databases (KDD)

methodology; (b) Cross-industry standard process for data mining (CRISP-DM) meth-
odology.

You must provide an explanation of how the raw data was gathered/compiled and
analysed. Describe pre-processing and the statistical techniques used upon the data.
Detail all the steps from data collection to final results. Write a subsection for each
step.
4
https://ojs.aaai.org/aimagazine/index.php/aimagazine/article/download/1230/1131
5
https://en.wikipedia.org/wiki/Cross-industry_standard_process_for_data_mining

7
4 Design Specification (overview of new algorithm
or model being proposed)
If a new algorithm or model is proposed, a word based description of the algorithm/model
functionality should be included here. A drawing of the new neural network, classifier, or
proposed data flow should be included here. Contrast the techniques and/or architecture
and/or framework that underlie the implementation and the associated requirements. No
implementation or actual code is required here. A screenshot of code could be included
on the Configuration Manual, if required, while implementation details will appear in
Section 5.

5 Implementation
You will of course want to discuss the implementation of the proposed solution. Only the
core part of the implementation should be described.
It should describe the outputs produced, e.g. transformed data, code written, models
developed, questionnaires administered. The description should also include what tools
and languages you used to produce the outputs. This section must not contain code
listing or user manual description.

Hints
✘ Do not copy/paste code or screenshots from your program output. It looks ugly!
✓ Include a diagram with the logical flow of your implementation (zoom in on the one
used for KDD/CRISP-DM in Section 3.

6 Evaluation (experiments & results)

Here is where you show what you did and what you got, and then what does all these
mean. The first section is about the experiments, then comes the comprehensive analysis
of results and main findings as well as the implications of these findings:
• in subsections Experiment / Case Study you basically write the rationale to
create the experiment, what do you plan to evaluate or look at, and then write
what you got.
• while in Discussion you take each Experiment / Case Study and interpret the
results, one-by-one.

Hints
✘ Too small font sizes make plots difficult to read and interpret.
✓ Only include material aligned to the answer of your research question.

6.1 Experiment / Case Study 1

did X got Y . . .

8
6.2 Experiment / Case Study 2
did X got Y . . .

6.3 Experiment / Case Study 3

did X got Y . . .

6.4 Experiment / Case Study N

did X got Y . . .

6.5 Discussion (interpretation or what those results mean)

We finally arrive at the core part of your document. You should essentially provide
an interpretation of your results. A figure or table, such as the one shown in Table 3,
summarising the experiments / case studies would definitively help you.

Table 3: Summary of findings

Experiment Main findings and discussions

experiment1 got Y1 ; Why did you get that? Is it expected? What does it mean?
experiment2 got Y2 ; Why did you get that? Is it expected? What does it mean?
experiment3 got Y3 ; Why did you get that? Is it expected? What does it mean?
experimentN got YN ; Why did you get that? Is it expected? What does it mean?

Include a detailed discussion of the findings from each Experiment / Case study,
one-by-one. At least one paragraph for each Experiment / Case study. Hints: what
does it mean that loss function? does it converge? what does it mean that histogram?
what does it tell me that confusion matrix? Do not forget to explain each figure and
table you include in the document. Think in the “so-what” question and put yourself in
the boots of a hypothetical reader, someone outside the College in the years to come.

Hints
✘ do not cherry-pick results that confirms your existing beliefs or ideas (avoid con-
firmation bias)

✓ be honest, it is fine to say something did not work; negative results are results.

✓ take the summary from Table 3 and provide an in-depth, rigorous analysis of results
(one-by-one).

Stay focused and use results to create a “story”; only the most relevant results related
to your research question and objectives shall be presented. Statistical tools should be
used to critically evaluate and assess the experimental research outputs and levels of
significance. Use visual aids such as graphs, charts, plots and so on to show the results.

9
Hints
✘ Do not be shy. Write! Here is where you showcase your understanding of the
problem at hand.
✓ Make it abundantly clear that you know what the results mean and provide answers
the “so-what?” question6 .
Discuss your limitations. You should criticize the experiment(s), and be honest
about whether your design was good enough. Suggest any modifications and improve-
ments that could be made to the design to improve the results. You should always
put your findings into the context of the previous research that you found during your
literature review

7 Conclusion and Future Work (final recap)

Finally, a quick recap. Most readers will only go through your Abstract and Conclusion,
so make them self-contained.
• Restate your research question(s), and your objectives.
• Briefly mention your key findings, and state how successful you have been in an-
swering the research question(s).
• The research gap has hopefully been narrowed, so what could you do to bridge the
gap even further?
• Write a couple of paragraphs about any recommendations for future work or po-
tential for commercialisation.

Hints
✘ Do not introduce new results here. Conclusions should mainly draw from Sec-
tion 6.5.
✘ Do not include figures or tables here.
✓ What can you do now with your results? what is the main learning?
✓ Write about something you wish you had done differently
Present meaningful future work. Sweeping more parameters in your simulation /
model / platform is probably not meaningful. Discuss what could a follow up research
project be. What would you recommend to improve or try differently or expand upon.

Note:
It is time to write your Abstract from pieces of Discussion and Conclusions. It is im-
portant to tell readers why your work is so important, why should we all care about it,
what is the relevance of this work? Expected key sentence are: We found / discovered /
analysed / confirmed / proposed a novel X / etc.
6
Google it or read the following paper https://doi.org/10.1080/17439884.2013.848454

10
References
Beloglazov, A. and Buyya, R. (2015). Openstack neat: a framework for dynamic and
energy-efficient consolidation of virtual machines in openstack clouds, Concurrency and
Computation: Practice and Experience 27(5): 1310–1333.

Feng, G. and Buyya, R. (2016). Maximum revenue-oriented resource allocation in cloud,

IJGUC 7(1): 12–21.

Gomes, D. G., Calheiros, R. N. and Tolosana-Calasanz, R. (2015). Introduction to the

special issue on cloud computing: Recent developments and challenging issues, Com-
puters & Electrical Engineering 42: 31–32.

Kune, R., Konugurthi, P., Agarwal, A., Rao, C. R. and Buyya, R. (2016). The anatomy
of big data computing, Softw., Pract. Exper. 46(1): 79–105.

11