Unsupervised Learning is a class of problems or algorithms in Machine Learning where unstructured data is given as input with the goal of finding more structure in the data output. The word “Structure” can reference whether the input data includes a “right” answer, label, or grouping.
In an example, you may be given health data of several patients sharing a common illness. The health data are the variables, and the common illness is the relationship. Using this data, you may be able to predict whether a patient has this common illness. This is an example of Supervised Learning. Specifically a classification problem.
Instead of providing the bit of information about the patients’ shared common illness relationship, you instead attempted to use the health data to determine if there were any structure, commonalities, or relationships which can be found. This could mean a learning algorithm would output patients into 2 groupings: Group 1 and Group 2. The algorithm is not aware of what this grouping means, but it was able to notice there was enough of a similarity with patients in each group to result in two distinct groupings were recognized. This is an example of Unsupervised Learning. Specifically a clustering problem.