Matt Scholar Solutions

Computer science and data science

Dissertation mentoring for computer science and data science students

Build the research and technical skills behind a strong computing dissertation. Get practical guidance on methodology, Python, machine learning, deep learning and evaluation, then apply what you learn to your own work.

Research design

Typical kinds of computing and data science dissertation

Computing dissertations can take several forms. The right design depends on your research question, available data and programme requirements. Knowing the shape of the project helps you identify the skills and evaluation plan you will need.

TypeWhat it involvesSkills you’ll usually need
Machine-learning model developmentBuild and compare models for a clearly defined prediction or classification task, using suitable data and a reproducible training process.Python · scikit-learn · feature engineering · model evaluation
Deep-learning applicationApply a neural-network approach to a bounded task such as image classification, text analysis or time-series prediction, and justify the architecture and training choices.Python · TensorFlow or PyTorch · training diagnostics · evaluation
Comparative algorithm studyCompare selected algorithms under consistent data preparation, validation and evaluation conditions, then explain where their performance differs.Research design · baselines · cross-validation · statistical comparison
Data analysis and visualisationUse a relevant dataset to answer a focused question through careful cleaning, exploratory analysis, appropriate statistical methods and clear visual communication.Data preparation · quantitative analysis · data visualisation
Systematic literature reviewIdentify, screen and synthesise research on a focused computing question with a documented search and selection process.Research methodology · systematic literature review · critical synthesis
Software or system evaluationDesign and evaluate a software artefact, tool or system against explicit requirements using a suitable test plan, benchmark or user evaluation.Research design · implementation plan · evaluation metrics · reproducibility

Check your dissertation handbook and speak with your supervisor before committing to a design. Requirements for implementation, data access, ethics and evaluation vary between programmes.

Practical guidance

Common questions from computing and data science students

Technical choices make more sense when they follow from the research question and the evidence you can obtain. These are some of the questions students commonly work through.

How do I choose a machine-learning model for my dissertation?

Start with the research question, target variable and available data. Define a simple baseline, then compare a small number of justified alternatives using the same split or cross-validation plan and metrics. A more complex model is not automatically a better research choice; explain why each method is suitable and what its limitations are.

How should I evaluate a classification model?

Choose metrics that reflect the task and the costs of different errors. Accuracy can hide poor performance on an imbalanced dataset, so consider precision, recall, F1 score, a confusion matrix and, where appropriate, ROC-AUC or precision-recall analysis. Keep a final test set separate from model selection and describe the validation process clearly.

Can I use a public dataset?

Often, yes, provided it is relevant to a focused question, accessible under appropriate terms and suitable for your planned analysis. Check its provenance, collection context, missingness, class balance, possible leakage and licence. Explain how the data was selected and prepared, and follow your university's ethics and data-handling requirements.

Do I need to build a new algorithm?

Usually not. A dissertation can make a sound contribution by applying and evaluating established methods in a well-motivated context, comparing approaches fairly, or analysing a problem carefully. Your contribution should be explicit and appropriate to your programme's requirements; discuss expectations with your supervisor.

Should I use TensorFlow or PyTorch?

Choose a framework that fits your task, prior experience, available examples and programme requirements. Both TensorFlow and PyTorch can support deep-learning projects. Keep the implementation reproducible and explain the model, training setup and evaluation choices rather than treating the framework itself as the research contribution.

Is Google Colab suitable for dissertation work?

Google Colab can be a practical environment for learning and running notebook-based experiments, subject to its current resource limits and your institution's data rules. Keep a local or version-controlled copy of your notebooks and record dependencies, data sources, random seeds and settings so you can reproduce the analysis.

Model evaluation

How do you compare machine learning models statistically?

Accuracy alone may not show which model is more useful. It can conceal poor performance on a minority class or differences in the kinds of errors models make. Choose metrics that fit your research question, the outcome and the balance of your dataset.

Compare models under the same evaluation protocol: use consistent data preparation, splits, validation strategy and metrics. For a paired comparison, the predictions or scores should come from comparable observations, such as the same held-out test cases or matching validation folds. This makes the comparison fairer and helps avoid treating unrelated results as if they were paired.

Comparing predictive performance is not the same as establishing statistical significance. A score difference describes observed performance; a significance test assesses how compatible that difference is with chance under a particular design and its assumptions. The suitable test depends on the experimental design, number of models and evaluation setup, so no single test applies to every dissertation.

Interpret significance alongside practical importance. Where appropriate, report effect sizes, uncertainty and limitations, including the dataset size, validation choices and variability across runs. A small difference can be statistically detectable without making one model meaningfully better for the intended use.

Start with a feasible question

Choosing your topic

A workable computing topic is focused enough to investigate, feasible with your time and skills, and supported by data or a system you can access. Be clear about the problem, the intended contribution and how you will evaluate your work before you commit to a method.

Use the complete topic guide to narrow your idea, check data access and plan a suitable evaluation.

Read the Computer Science and Data Science topic guide

From idea to defence

Your path by stage

The work moves through connected decisions. Each stage produces something useful for the next one.

  1. 01

    Choose a topic

    Narrow an interest into a question that fits your timeline and available evidence.

    Research Topic and Proposal Mentoring
  2. 02

    Plan the research

    Set a method, data plan and evaluation criteria that answer the question.

    Research methodology
  3. 03

    Prepare the data

    Check provenance, clean and explore data, and make the preparation reproducible.

    Quantitative analysis
  4. 04

    Implement and evaluate

    Build a baseline, run fair comparisons and interpret results against your criteria.

    Python and Machine Learning
  5. 05

    Present and defend

    Explain your choices, findings, limitations and contribution with confidence.

    Mock Viva and Presentation Coaching

More to know

FAQ

What can I bring to a free consultation?

Bring your course and deadline, the question or topic you are considering, any supervisor feedback, and the tools or methods you expect to use. If you have a brief or handbook, its requirements help us discuss a realistic next step.

Do you write code or dissertation chapters for students?

No. Mentoring focuses on helping you understand the research and technical skills so you can make decisions, write your own work and implement your own analysis. We can explain methods, work through examples and help you learn how to troubleshoot.

Can you help if I have already started my project?

Yes. Students come at different stages, from topic selection and proposal planning through data preparation, implementation, evaluation and viva preparation. A consultation can identify the specific skill gap and whether mentoring fits your needs.

How are mentoring sessions delivered?

Sessions are live online and planned around your stage and learning goals. Practical demonstrations use examples and suitable sample data; you then apply what you learn independently to your own project.

Which programming tools do you teach?

The tools depend on the project. Computing and data-science mentoring may cover Python, scikit-learn, TensorFlow, PyTorch and Google Colab, alongside research methodology and evaluation. We agree the relevant tools in the consultation.

One clear next step

Talk through your dissertation with a mentor

Tell us your course, project stage and where you feel stuck. We’ll help you identify a practical learning plan in a free consultation.

Written by the Matt Scholar research team · Last reviewed 17 November 2026

References

  • Bishop, C. M. (2006). Pattern Recognition and Machine Learning. New York: Springer.
  • Goodfellow, I., Bengio, Y. and Courville, A. (2016). Deep Learning. Cambridge, MA: MIT Press.
  • James, G., Witten, D., Hastie, T. and Tibshirani, R. (2021). An Introduction to Statistical Learning: with Applications in R. 2nd edn. New York: Springer.
  • Sculley, D. et al. (2015). ‘Hidden technical debt in machine learning systems’. In: Advances in Neural Information Processing Systems, 28.
  • Wohlin, C. et al. (2012). Experimentation in Software Engineering. Berlin: Springer.