Applied machine learning work covering regression, classification, neural networks, and final assessed work on fraud detection under extreme class imbalance, developed across module exercises, a mid-module project, and a larger final project.
The project work built from earlier regression and classification notebooks into larger assessed machine learning pieces. The mid-module work used structured datasets such as heart disease and related tabular data, while the final project focused on credit-card fraud detection with an explicitly cost-sensitive evaluation strategy.
The final assignment compared class-weighted logistic regression with neural-network approaches including standard BCE, focal loss, and weighted BCE on a highly imbalanced transaction dataset. That work also included ROC analysis, reliability diagrams, class-distribution figures, and calibration-oriented reporting rather than relying on accuracy alone.
The module record shows a progression from introductory regression and classification material into neural-network work, then into larger assessed notebooks and formal writeups. The project folders include interim notebooks, saved figures, assignment PDFs, reference lists, and final dissertation documents, which is why this section is better described as a substantial body of work rather than a single isolated notebook.