Exploring Patterns in Nuclear Physics Data Through Machine Learning

This thesis explores the application of machine learning algorithms to nuclear physics data,
aiming to uncover patterns within the data and reveal relationships between various nuclear
characteristics. While the research demonstrated some success, particularly in identifying
correlations between separation energies, shell models, and magic numbers, it also
encountered significant challenges. The most significant among these was the limitation posed
by the quality and quantity of available data, which affected the accuracy and reliability of
predictions, such as those for proton and neutron drip lines.
The research adopted a broad, exploratory approach, intentionally avoiding the use of
established physics models to allow machine learning to independently identify patterns.
However, this wide-ranging focus, combined with data limitations, resulted in findings that are
insightful but often inconclusive. The experiments conducted, including attempts to relate
nuclear deformity to stability and to apply machine learning to a model influenced by the
polyspheron model, further underscored the need for better and more targeted data.
This thesis highlights the potential of machine learning in nuclear physics but also emphasises
the importance of depth and data quality in future research. The results provide a foundation
for more focused studies, where improved datasets and a narrower research scope could yield
more definitive insights.