The system understands the significance of the data points and a cognitive sub-process clusters the data, removing the need for a human supervisor.

A neural network is deployed in unsupervised mode to determine to which cluster a data point belongs.

Examples of data points and their associated cluster are fed into the system allowing it to learn and generalize for the rest of the data set.

Data points are compared at random and the differences between their features are used to determine whether they belong in the same cluster or not.