Abstract:
Gestational Diabetes Mellitus (GDM) is a type of diabetes that develops during pregnancy
and can affect both the mother and the fetus. Screening and identification of
GDM at an early stage is very vital to ensure that no complications take place. This thesis
proposes a framework for the early prediction of GDM employing machine-learning
techniques with particular adherence to model interpretability. The used methodology
incorporates data preprocessing, feature selection, evaluation of the model, and explainability
of the performed predictions. Thus, in synthesizing the study, the aspect
of Explainable AI ‘XAI’ is incorporated, guaranteeing the model gives accurate and
plausible predictions. In addition, our method demonstrates a better performance than
previous works even when we remove some of the features, stressing the significance of
the feature selection in increasing the effectiveness of the models. The explainability
methods increase the trust of users in the model predictions because the information is
made available to them hence improving their decision-making processes. It is in this
regard that the following thesis lays down the groundwork towards attaining a robust
and explainable system for early GDM identification. It applies in enhancing meternal
care and the baby through early detection and intervention and enhancing decisions
on healthcare.
