Unsupervised learning is about analyzing the data and discovering hidden structures in unlabeled data. As no notion of the right labels is given, there is also no error measure to evaluate a learned model; however, unsupervised learning is an extremely powerful tool. Have you ever wondered how Amazon can predict what books you'll like? How Netflix knows what you want to watch before you do? The answer can be found in unsupervised learning. The following is one such example.
Find similar items
Many problems can be formulated as finding similar sets of elements, for example, customers who purchased similar products, web pages with similar content, images with similar objects, users who visited similar websites, and so on.