This process calculates the average values for all attributes grouped by the class (label). The resulting prototypes or centroids are then used for a nearest neighbor model which is applied again on the full data set for demonstration purposes.