Abstract
Medical data generated from hospitals are an increasing source of information for automatic medical diagnosis. These data contain latent patterns and correlations that can result in better diagnosis when appropriately processed. Most applications of machine learning (ML) to these patient records have focused on utilizing the ML algorithms directly, which usually results in suboptimal performance as most medical datasets are quite imbalanced. Also, labelling the enormous medical data is a challenging and expensive task. In order to solve these problems, recent research has focused on the development of improved ML methods, mainly preprocessing pipelines and feature learning methods. This thesis presents four machine learning approaches aimed at improving the medical diagnosis performance using publicly available datasets...
D.Ing. (Electrical and Electronic Engineering)