What form of learning involves grouping data points based on shared attributes without prior classification?

Clustering is a type of unsupervised learning that focuses on organizing a set of data points into groups, or clusters, based on shared attributes or similarities among them. In clustering, the algorithm does not require pre-labeled data or prior classification; rather, it identifies patterns and structures within the data itself.

The primary goal of clustering is to discover inherent groupings in data. For instance, it can be used in market segmentation, where customers with similar buying behaviors are grouped together, enabling businesses to tailor their offerings accordingly. This technique is widely applicable in various domains, including image analysis, social network analysis, and customer segmentation.

In contrast, feature learning involves transforming raw data into a more informative representation, which may eventually aid in supervised learning tasks but does not inherently involve grouping data. Prediction and regression are types of supervised learning that rely on labeled data to forecast outcomes based on input features, diverging from the principle of clustering where no labels are provided beforehand.