STFC Hartree Centre

Daniel Rodriguez examines how the structure of tabular datasets affects classification performance and the stability of feature selection, using software-defect prediction as a case study. The work evaluates data-complexity metrics across multiple algorithms to understand the effects of dimensionality, class overlap, and network structure. These relationships inform the choice of predictive models and feature-selection strategies suited to particular datasets, with the aim of producing more reliable and interpretable defect-prediction systems.