A machine learning approach that finds patterns in unlabeled data. Without predefined outcomes, unsupervised algorithms identify structure—such as clustering or association—solely from the input features. In trading, unsupervised learning can uncover hidden relationships (e.g. grouping similar currency behaviors or reducing dimensionality of indicator sets). Techniques include clustering (like k-means) and dimensionality reduction (like PCA). Unsupervised methods can help a forex robot detect regimes (bull vs bear markets) or create input features automatically, aiding strategy development when explicit labels are unavailable.

Back to Robots Glossary Index