Introduction to Decision Trees - Data Mining with Decision Trees: Theory and Applications - page 12

be used by an intelligent system. Knowledge elicitation is usually dicult
because it is challenging to find an available expert who would be willing
to provide the knowledge engineer with the information he or she needs to
create a reliable expert system. In fact, the diculty inherent in the process
is one of the main reasons why companies avoid intelligent systems. This
phenomenon constitutes the knowledge elicitation bottleneck.
A decision tree can be also used in order to analyze the payment ethics
of customers who received a mortgage. In this case there are two classes:

Paid (denoted as “P”) — The recipient has fully paid off his or her
mortgage.

•

Not Paid (denoted as “N”) — The recipient has not fully paid off his or
her mortgage.

•

This new decision tree can be used to improve the underwriting decision
model presented in Figure 16.1. It shows that there are relatively many
customers who pass the underwriting process but that they have not yet
fully paid back the loan. Note that as opposed to the decision tree presented
in Figure 16.1, this decision tree is constructed according to data that
was accumulated in the database. Thus, there is no need to manually
elicit knowledge. In fact, the tree can be built automatically. This type of
knowledge acquisition is referred to as knowledge discovery from databases.
The employment of a decision tree is a very popular technique in data
mining. Many researchers argue that decision trees are popular due to their
simplicity and transparency. Decision trees are self-explanatory; there is
no need to be a data mining expert in order to follow a certain decision
tree. Usually, classification trees are graphically represented as hierarchical
structures, which renders them easier to interpret than other techniques. If
the classification tree becomes complicated (i.e. has many nodes) then its
straightforward graphical representation become useless. For complex trees,
other graphical procedures should be developed to simplify interpretation.

1.8 Characteristics of Classification Trees

A decision tree is a classifier expressed as a recursive partition of the
instance space. The decision tree consists of nodes that form a Rooted
Tree, namely, it is a Directed Tree with a node called a “root” that has no
incoming edges. All other nodes have exactly one incoming edge. A node
with outgoing edges is referred to as an “internal” node or a “test” node.
All other nodes are called “leaves” (also known as “terminal” nodes or