Rule Builder > Decision Tables and Trees > Decision Table or Decision Tree?

The choice of decision table versus decision tree depends mainly on the uniformity of your data. If your data is such that the set of conditions for each rule is almost the same, with the same order, a decision table is the best choice. If the set of conditions is different in cardinality, in order, and kinds of condition, then a tree is a better solution.

In a decision table, uniformity implies fewer columns, all or most of which have cells with values, whereas non-uniformity means many columns, a lot of which have cells containing no values. This results in an untidy decision table that makes it difficult to read and difficult to identify the rules. Non-uniformity of data has little or no impact on a decision tree and in this case makes it a more suitable choice.

Figure 11.5 shows a non-uniform decision table. In the table there are many columns and empty cells and it is not easy to follow the path of a given rule. The table is unclear and difficult to use.

images/dt_asymmetric.png

Figure 11.5    Non-uniform Decision Table

Figure 11.6 shows the same decision table converted to a decision tree. The decision tree is much more usable and easy to follow. Clicking on the leaf of a branch reveals the rule path in blue.

images/dtree_asymmetric.png

Figure 11.6    Non-uniform Decision Table Displayed as a Decision Tree

See Also

Spreadsheet View


Customer Support | Copyright © 1987-2004 ILOG S.A. All rights reserved. Legal terms. PREVIOUS   NEXT