Package adaa.analytics.rules.logic.induction

Class ClassificationExpertFinder

java.lang.Object

adaa.analytics.rules.logic.induction.AbstractFinder

adaa.analytics.rules.logic.induction.ClassificationFinder

adaa.analytics.rules.logic.induction.ClassificationExpertFinder

All Implemented Interfaces:

IExpertFinder, AutoCloseable

public class ClassificationExpertFinder
extends ClassificationFinder
implements IExpertFinder

Class for growing and pruning classification rules with user's knowledge.

Field Summary

Fields

Modifier and Type	Field	Description
protected Knowledge	knowledge	User's knowledge.

Fields inherited from class adaa.analytics.rules.logic.induction.ClassificationFinder

precalculatedCoverings, precalculatedCoveringsComplement

Fields inherited from class adaa.analytics.rules.logic.induction.AbstractFinder

attributeValuesOrder, modifier, params, pool, threadCount

Constructor Summary

Constructors

Constructor	Description
ClassificationExpertFinder(InductionParameters params, Knowledge knowledge)	Invokes base class constructor and initializes knowledge object.

Method Summary

Modifier and Type	Method	Description
void	adjust(Rule rule, IExampleSet dataset, Set<Integer> uncoveredPositives)	Adjusts expert rule.
protected boolean	checkCandidate(ElementaryCondition cnd, double classId, double p, double n, double new_p, double P, double uncoveredSize, int ruleOrderNum)	Checks if candidate condition fulfills coverage requirement and is not in conflict with forbidden knowledge.
int	grow(Rule rule, IExampleSet dataset, Set<Integer> uncoveredPositives)	Adds elementary conditions to the classification rule premise until termination conditions are fulfilled.
void	setKnowledge(Knowledge knowledge)	Sets users knowledge.

Methods inherited from class adaa.analytics.rules.logic.induction.ClassificationFinder

induceCondition, postprocess, preprocess, prune, tryAddCondition

Methods inherited from class adaa.analytics.rules.logic.induction.AbstractFinder

addObserver, clearObservers, close, names2attributes, notifyConditionAdded, notifyConditionRemoved, notifyGrowingFinished, notifyGrowingStarted, notifyRuleReady

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Details

knowledge

protected Knowledge knowledge

User's knowledge.

Constructor Details

ClassificationExpertFinder

public ClassificationExpertFinder(InductionParameters params,
 Knowledge knowledge)

Invokes base class constructor and initializes knowledge object.

Parameters:

params - Induction parameters.

knowledge - Expert knowledge.

Method Details

setKnowledge

public void setKnowledge(Knowledge knowledge)

Sets users knowledge.

Parameters:

knowledge - User's knowledge to be set.

adjust

public void adjust(Rule rule,
 IExampleSet dataset,
 Set<Integer> uncoveredPositives)

Adjusts expert rule. The procedure specifies value sets of adjustable conditions, i.e., those in the form:

• attribute ~= Any,
• attribute ~= SomeSet (non-empty intersection of determined value set with SomeSet is required).

Specified by:

adjust in interface IExpertFinder

Parameters:

rule - Rule to be adjusted.

dataset - Training dataset.

uncoveredPositives - Set of positive examples not covered by the model.

grow

public int grow(Rule rule,
 IExampleSet dataset,
 Set<Integer> uncoveredPositives)

Adds elementary conditions to the classification rule premise until termination conditions are fulfilled. The method uses expert knowledge.

Overrides:

grow in class ClassificationFinder

Parameters:

rule - Rule to be grown.

dataset - Training set.

uncoveredPositives - Set of positive examples yet uncovered by the model.

Returns:

Number of conditions added.

checkCandidate

protected boolean checkCandidate(ElementaryCondition cnd,
 double classId,
 double p,
 double n,
 double new_p,
 double P,
 double uncoveredSize,
 int ruleOrderNum)

Checks if candidate condition fulfills coverage requirement and is not in conflict with forbidden knowledge.

Overrides:

checkCandidate in class ClassificationFinder

Parameters:

cnd - Candidate condition.

classId - Class identifier.

Returns: