Tree terminology

Tree¶

The Tree structure represents a hierarchical segmentation of the base selection. At each branch in the tree, the base selection is split into tow sub-selections.

Each node is a sub-selection of the full base selection. A node can be said to "contain" a number of people/products etc (depending on the resolve level of the selections) from the full base selection.
Since the analysis selection is a subset of the base (e.g. the people who went to Sweden), each node will contain a sub-selection of the full analysis selection.
The purpose of a Decision Tree is to create nodes which have a particularly high or particularly low proportion of people from the analysis selection (the Analysis %).

The root node is the starting node of the tree and contains the full base selection.

A number of child nodes are formed whenever a node is split. All the records within a node that is split are allocated to one of the child nodes.
There will be a selection rule associated with each child node (e.g. “Income <£20k”) which describes how this child is formed from its parent.

The node rule is the selection rule associated with each child node (e.g. "Income <20k").
Strictly speaking, this describes how the child is formed from its parent (last rule).
In practice, the full rule is more useful, as this describes how to form a node starting from the root node. It is simply the individual rules from each branch node along the path followed to reach the node in question.

A parent node is one that has been split from child nodes. All the records within a node that is split are allocated to one of the child nodes.

A leaf node is one that is not split further. It is therefore at the end of a series of branch nodes.

The level or depth of a particular node relates to how many splits are needed to get from the root node to the node in question.
The root node is defined as being level 1 and depth 1.