Association Rules - Rule building
Items per rule
Use these options to specify how many items, or values, can be used in each rule.
Note: The
combined total of these two fields cannot exceed 10.
- Maximum conditions
- Select the maximum number of conditions that can be included in a single rule.
- Maximum predictions
- Select the maximum number of predictions that can be included in a single rule.
Rule building
Use these options to specify the number and type of rules to build.
- Maximum number of rules
- Specify the maximum number of rules that can be considered for use in the building of rules for your model.
- Rule criterion for top N
- Select the criterion that is used to establish which are the top N rules, where N is the value
that is entered in the Maximum number of rules field. You can choose from the
following criterion.
- Confidence
- Rule Support
- Condition Support
- Lift
- Deployability
- Only true values for flags
- When your data is in tabular format, select this option to include only true values for flag fields in the resulting rules. Selecting true values can help make rules easier to understand. The option does not apply to data in transactional format. For more information, see Tabular versus Transactional Data.
Rule criterion
If you select Enable rule criterion, you can use these options to select
the minimum strength that rules must meet to be considered for use in your model.
- Confidence Specify the minimum percentage value for the Confidence level for a rule that is produced by the model. If the model produces a rule with a level less than this amount, the rule is discarded.
- Rule Support Specify the minimum percentage value for the Rule Support level for a rule that is produced by the model. If the model produces a rule with a level less than this amount, the rule is discarded.
- Condition Support Specify the minimum percentage value for the Condition Support level for a rule that is produced by the model. If the model produces a rule with a level less than the specified amount, the rule is discarded.
- Lift Specify the minimum Lift value that is allowed for a rule that is produced by the model. If the model produces a rule with a value less than the specified amount, the rule is discarded.
Exclude rules
In some cases, the association between two or more fields is known or is self-evident, in such a case you can exclude rules where the fields predict each other. By excluding rules that contain both values, you reduce irrelevant input and increase the chances of finding useful results.
- Fields
- Select the associated fields that you do not want to use together in rule building. For example, associated fields might be Car Manufacturer and Car Model, or School Year and Age of Pupil. When the model creates rules, if the rule contains at least one of the fields selected on either side of the rule (condition or prediction), the rule is discarded.