# How to mitigate the hierarchical error propagation in tree-structured classification

Suppose we have a multi-class classification problem, where the number of classes $K \geq 3$

We use a tree structure of multiple SVMs to divide and conquer the problem, with one example in the figure below.

Generally, in this kind of hierarchical classification, if a sample is wrongly classified in an upper layer, it is less likely to be corrected again in lower layers.

My question: Is there any strategy to mitigate such hierarchical error propagation in tree-structured classification?

Much appreciated!

• I'm not sure I accept the premise of the problem you are trying to solve. If an object is misclassified in a high level, then it is certainly misclassified in a lower level: if classifying a book as a food is wrong, then classifying a book as a fruit is certainly wrong, whereas you suggest that such misclassification may be corrected in lower levels. Does your hierarchical relation have the property all fruits are food, or is the relationship more fuzzy than that? – Lieuwe Vinkhuijzen May 4 '16 at 13:03
• Why do you limit yourself to tree-structured classification? If you don't like the disadvantages of tree-structure classification, the obvious solution is to look at a different method of doing multi-class classification (there are others that don't use a tree structure). What requirement do you have, that a non-tree-structured classification fails to meet? – D.W. May 4 '16 at 17:54