Unsupervised learning trains a model on data that has no labels, leaving the algorithm to discover structure on its own. The “Machine Learning Basics” chapter of Goodfellow, Bengio, and Courville’s “Deep Learning” describes unsupervised learning as observing several examples of data and learning useful properties of its structure, in contrast to supervised learning where each example has a target label.
Typical unsupervised tasks include clustering, which groups similar items together, and dimensionality reduction, which compresses data while keeping its important structure. For example, an unsupervised algorithm might segment customers into natural groups based on purchasing behavior without anyone defining the groups in advance.
Because no labels are required, unsupervised methods can exploit large pools of raw data. The trade-off is that the patterns they surface are not guaranteed to match any specific business goal and often need human interpretation.
Why business readers should care: Unsupervised learning is valuable when you have lots of data but few labels, such as exploring customer segments or spotting anomalies. It can reveal patterns you did not know to look for, but its outputs usually require expert review before acting on them.