What characterizes unsupervised learning?

Unsupervised learning is characterized by its ability to discover hidden structures within the data without the need for labeled datasets. In this approach, algorithms analyze unlabelled data to identify patterns, groupings, or relationships that may not be immediately apparent. This is beneficial when the goal is to explore the data and extract insights, as it allows for the identification of natural clusters or anomalies in the dataset.

Unlike supervised learning, which relies on labeled examples where the outcome is known, unsupervised learning does not have predefined labels and instead seeks to understand the underlying structure on its own. This makes it particularly useful for exploratory data analysis, as well as applications like customer segmentation, anomaly detection, and recommendation systems.

The other options imply the necessity of labels or external guidance, which does not align with the nature of unsupervised learning. Specifically, the requirement for labeled datasets, external guidance, and the focus on classification tasks all pertain to supervised learning methods rather than unsupervised learning.