Novel Methods for Positive Unlabeled Learning in Imbalance Data Sets with Applications in Fraud Detection KU Leuven
Most modern machine learning techniques require sufficiently large labeled training data to achieve good performance. However, several real-world applications suffer from incomplete label information, such as recommendation systems, medical diagnosis, bioinformatics, and fraud detection. The acknowledgment of this issue of incomplete labels led to more attention towards weakly supervised learning. This dissertation focuses on the specific ...