|
|
Rule Node |
The rule node builds, imports or exports a set of one or more rules in conjunctive form. The rules can be evaluated against the database and the results displayed on screen, or used to filter records meeting a set of conditions.
Rules that have been generated from the discovery node (SA) and copied to the scratchpad can be pasted back into a rule node for evaluation. Alternatively, you can create rules by using the graphical rule builder. The choice not to allow free format creation of rules helps avoid parse errors that commonly occur in packages and act as an annoyance to the user.
One possible use of the rule node is to "tidy up" rules generated from the discovery node and perform a sensitivity analysis on the results. For example, the rule "Satisfaction > 5.9765" is quite precise and for neatness could be changed to "Satisfaction > 6" if it makes little, or no, difference to the performance of the rule.
A special use of the rule node is to filter records according to certain conditions. You can construct a rule with no postcondition that can be used to filter records in or out of the stream. This feature is particularly useful when using the SA to generate multiple rules for a single class. After the SA has constructed the first rule, the records to which it applies can be filtered out using the rule node and the SA can be run again on the remaining records.
Rules can be evaluated on an individual basis or in disjunction with one another. When rules are evaluated, an HTML report shows you the properties of those rules, such as accuracy and coverage. You can view rules in either a proprietary format or using SQL.
Rules stored in a rule node can be exported in text or XML format for use in other applications that may wish to exploit the results of the KDD project.
For more information on rules see the definitions page.
Full details of the options available for the rule node can be found on the rule options page.
