Biomedical Engineering Reference
In-Depth Information
Decision Trees
Decision trees are hierarchically arranged questions and answers that lead to classification. As shown
in
Figure 7-12
, decision trees are formed by input nodes and tests on input data and whose leaf
nodes are categories of those data. In the decision tree in the figure, the tests (Test 1-Test 8) result
in textual categories (categories (A)-(H)), but they can also result in numerical categories. An
advantage of using decision trees in data mining is that they can be easily read and modified by
humans. For example, the results of Test 1 may lead to Test 2, 3, or 4. Once Test 2 is selected, the
only options are to characterize the input as belonging to category (A), or to select Test 7. Category
(A) can represent a particular family of proteins, for example. The only options from Test 7 are to
place the data into category (A) or category (F).
Figure 7-12. Decision Trees. A decision tree categorizes a pattern by
filtering it down through the tests in a tree.
