Decision Tree Induction

Learning process of a Decision Tree.

The simplest (and trivial) decision tree would be to have a path for each input example to a leaf node. This would perfectly realize the targets but would not generalize well on unseen data. Instead we prefer to find more compact trees by only looking at attributes that are the most important for deciding on the target value.

• Use an Attribute Selection Method to select an attribute that is best for splitting the tree into new branches
• Split the Tree
  • Use the new subtrees to repeat the steps
• Stop when
  • all tuples belong to the same class
  • all attributes have been used for splitting
  • no tuples has to be classified anymore
  • max node depth is reached
  • minimum number of instances in current node is reached
  • minimum impurity in current node is reached

CART

![[CleanShot 2023-09-30 at 11.31.31@2x.png]]