Unsupervise learning

Unsupervised learning aims to find patterns in data. The reason it is called unsupervised is that there is no label, classification or categorization to indicate the association of the features to the labels; so there is no supervision from the data.

The three most popular unsupervised learning methods include principal component analysis (PCA) clustering methods (k-mean and hierarchical) analysis and autoencoders.

Principal component analysis (PCA) , a self explanatory term, is a way to find the most principal components in a multi-dimensional data. In other words, PCA seeks what linear combination of different dimensions can encapsulate more information of the data.

Clustering create groups or segments that share attributes and similarities. This approach can help to identify the anomalies and anomalous data.

Autoencoders, is a method that in principle follows the same idea of the PCA but from a more general perspective. Like PCA, this approach aims to find the major component of a multidimensional data which can be a nonlinear combination of the dimensions.