January 2022 - June 2023:
|
Evaluating the impact of student-centred computational activities in Financial MathematicsWe conducted an evaluation of the computational lab component within a mandatory undergraduate Computational Finance course. The objective was to investigate the impact of the computational design of each lab on students' understanding of computational thinking. Additionally, we explored the relationship between financial mathematics and computational thinking for the students.
We analysed students' responses to weekly post-lab Google Surveys over a period of ten weeks in 2021. Thematic analysis was applied to analyse the students' responses and identify recurring patterns. From this analysis, we generated a set of 11 categories representing the main themes that emerged from the students' feedback. This project provided valuable insights into the effectiveness of the computational lab component in enhancing students' understanding of computational thinking within the context of financial mathematics. These findings contribute to the ongoing efforts to refine and improve the curriculum of the Computational Finance course, ensuring that students receive a comprehensive education in both financial concepts and computational skills. This work was published in the conference proceedings of the 9th Conference on Research in Mathematics Education in Ireland (MEI), The published conference paper is available here. Supervisors: Dr. Adamaria Perrotta and Dr. Anthony Cronin. |
June 2022 - August 2022: |
Application of ClinTrajan to UK Biobank Data |
Heart Failure is a clinical syndrome caused by structural and functional defects in the myocardium, which leads to impairment of ventricular filling or the ejection of blood. In 2020, 2.23% of people of 40 or more years of age had heart failure, making it a prevalent condition.
The main technique/method used in this project is ClinTrajan (Golovenkin, et al. 2020), the aim of which is to extract clinical (disease) trajectories from large-scale clinical datasets. The method is derived from single cell RNAseq trajectory analysis algorithms. The ClinTrajan method itself is implemented through a convenient Python package. Within the time constraints of this project, we successfully applied the ClinTrajan algorithm/method to UK Biobank (UKB) data, meaning this method can be used for future projects that are based on UKB data and we found three main disease trajectories:
Supervisor: Dr. Paul McGettigan (Novartis). |
October 2021 - January 2022: |
Aquaculture Meets Artificial Intelligence |
Fish farms are of great importance in Ireland, with 14 of Ireland's 19 coastal counties involved in fish farming, and the sector producing approx. €208 million worth of produce in 2017. Ireland has led the world in organically certified farmed fish for over 30 years.
It is known that if a fish farm experiences sufficiently high force (usually from storms), the net of the fish farm will break, the fish will escape and a lot of money will be lost. Typically, fish farms are equipped with several acceleration sensors; however, force sensors are very expensive and thus not easily accessible. This led us to our research question: Can the acceleration data from a fish farm be used to predict when the force measurements cross a threshold, damaging the farm? We spent 3 months on this project, using machine learning models to do our predictions. It was presented at the SFI CRT Winter Symposium in January, 2022. Group Members: CJ Clarke, Thiago Da Silva Cardoso, Shreyan Banerjee. Supervisors: Assoc. Prof. Vikram Pakrashi, Dr. Michelle Carey. |
June 2021 - December 2021: |
JSUparameters (R Package) |
My Master's thesis was entitled "How can we use QQ plots for estimation?" and the main output of this project was an algorithm that returns the parameter estimates of the best-fitting JohnsonSU distribution for a given dataset, whilst also considering all limiting cases of the distribution.
This algorithm is likely to be useful within the financial sector when investors are modelling investment returns and trying to make predictions. The JohnsonSU distribution is a much better fit for the distribution of stock returns than the normal distribution (see accompanying figure), which is typically used in modelling. This algorithm became publicly available on CRAN in an R package called "JSUparameters" in Dec. 2021; and, currently, it has over 6,500 downloads.
|