Unsupervised learning is a branch of machine learning that learns from test data that has not been labeled, classified or categorized. Unsupervised learning identifies common traits in the data and draws inferences based on the presence or absence of such commonalities in each new piece of data. Other branches of machine learning include supervised learning and reinforcement learning.
‘Labels’, in this context, refers to some outcome or target variable. When a target variable is not present in the data, there is no prediction for which to optimize.
In this case, unsupervised learning provides us with a number of exploratory data analysis algorithms to find hidden patterns or grouping in the data. Common applications of unsupervised learning include:
- Clustering - Identifying groupings of related observations
- Dimensionality Reduction - Reduces the number of variables taken into consideration via methods of feature selection and feature extraction
- Association Rule Learning - Discovering relationships between observations in a dataset