Hierarchical classification

Create a hierarchy of classification using PLS-DA, ML or SAM

The example above has two models: “Nuts or Shell” and “Type of Nut“ which are linked together in a hierarchical manner.

In this examples we use data from the Classification of nuts step 1 tutorials

The first model “Nuts or Shell” has a grouping of all types of nuts to a single group and shell as a separate classification, like so:

This is then used to create a initial sorting of objects as the Group “Nuts” or “Shell”.

We then add it to the Analyse Tree as a Classification of categories descriptor.

The second model “Type of Nut” is then created without the Shell class as it will not be applicable.

We exclude the Shell class from the model and train it as normal.

Then the model is added one level down from the “Nuts or Shell” classification node:

We connect the two nodes using the Connection parameter. We connect to the “Nuts” group. This will make it so the second model is only applied on the object if it classified as “Nuts” and the output of the second model is the which type of nut is actually is.