Semi-supervised learning

Halfway between the supervised and the non-supervised, partially supervised learning is based on mixed data in which a minimal part is already labeled and a very large majority consists of unlabeled data. This approach is used to improve the predictions made by the machine on unlabeled data and normally requires human intervention. The approach is mainly used in classification and clustering problems or in the description of the cause-effect relationships between variables.