Fig. 9: Mechanism of DT model, there is an initial tree and the data are input and go through each branch, the prediction carbonation depth Y’ will initially be obtained.
From: Machine learning for efficient CO2 sequestration in cementitious materials: a data-driven method

By comparing the loss between the real combination depth Y and Y', the decision tree will be further updated, e.g., by increasing the branches as indicated in the red frame. Finally, a final tree is supposed to get an acceptable prediction value.