BOOLRULE Procedure

Estimated Precision

Estimated precision helps BOOLLEAR shorten its search path and avoid generating overly specific rules. The precision is estimated by a form of additive smoothing with additional correction () to favor shorter rules over longer rules:

normal p normal r normal e normal c normal i normal s normal i normal o normal n Subscript i Superscript m Baseline left-parenthesis t right-parenthesis equals StartStartFraction normal upper T normal upper P Subscript i comma bold t Baseline plus StartFraction normal upper P Over normal upper N plus normal upper P EndFraction times m OverOver normal upper T normal upper P Subscript i comma bold t Baseline plus normal upper F normal upper P Subscript i comma bold t Baseline plus m EndEndFraction minus normal e normal r normal r Subscript i minus 1

normal e normal r normal r Subscript i Baseline equals StartFraction normal upper T normal upper P Subscript i comma bold t Baseline Over normal upper T normal upper P Subscript i comma bold t Baseline plus normal upper F normal upper P Subscript i comma bold t Baseline EndFraction minus StartStartFraction normal upper T normal upper P Subscript i comma bold t Baseline plus StartFraction normal upper P Over normal upper N plus normal upper P EndFraction times m OverOver normal upper T normal upper P Subscript i comma bold t Baseline plus normal upper F normal upper P Subscript i comma bold t Baseline plus m EndEndFraction plus normal e normal r normal r Subscript i minus 1

In the preceding equations, is a parameter that you specify for bias correction. A large m is called for when a very large number of rules are evaluated, in order to minimize selection bias. and are the true-positive and false-positive of rule when the length of the rule is i.

Last updated: August 06, 2026