Picking a Research Topic

Selecting a research topic is a daunting task. Here are some things to help you think through the process.

The first challenge of beginning your research study is that you need to find a topic you are interested in. Then you need to find a niche within that topic that still contains an open research problem.

Data Science

One line of thinking can always guide you when you pick a topic in data science. The first set of questions has to be:

  1. Have you got the data, or can you get it?
  2. If the data is publicly available. Has this already been done?
  3. If it has already been done. What is my contribution?

Machine Learning

In machine learning, the datasets for specific machine learning problems, the metrics for how comparisons between algorithms are done and the methodology for comparisons are generally well understood. Here, picking a topic should be more aligned with the area of machine learning you would like to contribute to.

Research Problem

Remember that even though a master’s unlike a PhD, need not be novel. It should still contribute to the field. Here is a quick discussion of the difference between the two.

Novelty

True novelty, for example, developing a groundbreaking new algorithm or theoretical framework, is challenging, even at the PhD level. Many seemingly novel ideas have already been explored in some form. The transition from a Master’s, where contribution is key, to a PhD, where novelty is often expected, can be a significant leap. Having already completed a Master’s, you understand the depth of existing research and how difficult it can be to make groundbreaking discoveries.

This challenge highlights the importance of:

  1. Thorough literature review,
  2. Creative thinking and
  3. Collaboration Mentorship.

Example: Developing a new type of neural network architecture that significantly outperforms existing models on a specific task.

Contribution

A Master’s dissertation focuses more on demonstrating a solid grasp of your field and your ability to conduct rigorous research. The emphasis is on showing that you can effectively utilise existing knowledge and tools to generate meaningful results and contribute to the ongoing discourse in your study area. A contribution is often achieved by advancing the field in some way, such as through:

  1. Applying existing methods to a new dataset or problem domain. This application expands the understanding of how these methods perform in different contexts.
  2. Improving existing methods. This improvement could involve refining an algorithm, optimising parameters, or combining techniques in a novel way to achieve better results.
  3. Conducting a thorough analysis and evaluation. This analysis could involve comparing different methods, identifying their strengths and weaknesses, or exploring the impact of different factors on performance.
  4. Generating new insights from data. Even if the methods used are not novel, the findings can be valuable.

Example: Applying a known machine learning model to predict customer churn in a specific industry, analysing the results and identifying key factors contributing to churn.