Unsupervised learning is a type of machine learning that involves training algorithms using data that is neither classified nor labeled. The system tries to learn the patterns and the structure from the data without any specific guidance. This contrasts with supervised learning, where the model is trained on labeled data. Unsupervised learning is particularly useful in identifying hidden patterns or intrinsic structures in data.