Boosting Accuracy of Classical Machine Learning Antispam Classifiers in Real Scenarios by Applying Rough Set Theory

Algorithm 1

Pseudocode of the proposed algorithm for the generation of RS revision rules.

(00) FUNCTION computeRules E: MessageIdentifierVector,

(01)             A: ConditionAttributeMatrix, X: DecisionAttributeMatrix);

(02)  X2: DecisionAttributeMatrix;

(03)  RED: AttributeSet;

(04)  R: Rule;

(05)  RESULT: Ruleset;

(06)

(07)  FOREACH  e INCLUDED IN E DO

(08)   FOREACH e INCLUDED IN E DO

(09)     IF (e == e ) THEN X2e = 1;

(10)     ELSE IF ( X == X ) THEN X2 = ?;

(11)     ELSE X2 = 0;

(12)   END_FOREACH;

(13)   RED = computeShortestReduct (E, A, X2);

(14)

(15)   FOREACH a INCLUDED IN R DO

(16)    IF (a INCLUDED IN RED) THEN

(17)      R.conditions = A, a;

(18)    ELSE R.conditions = ?;

(19)   END_FOREACH;

(20)   R.decision=X;

(21)   RESULT.add(R);

(22)  END_FOREACH;

(23)  RETURN RESULT;

(24) END_FUNCTION;