keel.Algorithms.Genetic_Rule_Learning.ILGA

Class SEM

java.lang.Object

keel.Algorithms.Genetic_Rule_Learning.ILGA.SEM

public class SEM
extends java.lang.Object

This class implements the SEM algorithm of the OIGA method, which evolves mono-attribute rules.

Constructor Summary

Constructors

Constructor and Description
SEM() Default constructor.
SEM(int pobSize, int numberRules, int attSel, InstanceSet IS) Parametrized constructor

Method Summary

Modifier and Type	Method and Description
void	crossOver() It performs a one point crossover in the new poblation, using adjacent chromosomes as parents
void	elitism() Copy the survivorsPercent proportion of the old poblation into the bottom half of the new one
void	evaluate() Its evaluate the NEW poblation, using a metric which summarizes the train CR and test CR
RuleSet	getChromosome(int i) Gets the i-th chromosome
double	getCR() Gets the Classification Rate of this SEM
void	mutate() Applies mutation in the new poblation
void	onePointCrossover(int cr1, int cr2) One-point crossover
void	run() It runs the Single-attribute Evolution Module (SEM) algorithm to obtain a rule set of ONE attribute
void	selection() Applies a roulette wheel selection
void	setGAparams(double mutationProb, double crossoverRate, double survivorsPercent) Set the parameters for this SEM
void	setGenerationLimit(int iters) Set the generations limit
void	setIS(InstanceSet dataset) Sets the reference data set for this SEM
void	tournament_selection() Applies a tournament selection, with tournament size of 2

Methods inherited from class java.lang.Object

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

Constructor Detail

SEM

public SEM()

Default constructor. No memory allocated

SEM

public SEM(int pobSize,
           int numberRules,
           int attSel,
           InstanceSet IS)

Parametrized constructor

Parameters:

pobSize - number of chromosomes (rule sets)

numberRules - number of rules of each chromosome

attSel - attribute used to evolve in this SEM

IS - the data set used for train the SEM

Method Detail

setIS

public void setIS(InstanceSet dataset)

Sets the reference data set for this SEM

Parameters:

dataset - new data set for training

setGenerationLimit

public void setGenerationLimit(int iters)

Set the generations limit

Parameters:

iters - maximum number of iterations of the SEM

setGAparams

public void setGAparams(double mutationProb,
                        double crossoverRate,
                        double survivorsPercent)

Set the parameters for this SEM

Parameters:

mutationProb - the mutation probability

crossoverRate - the crossoverProbability between 2 parents

survivorsPercent - the percent of parents that will be maintained from a generation to next one

getCR

public double getCR()

Gets the Classification Rate of this SEM

Returns:

retuns the Classification Rate of this SEM

getChromosome

public RuleSet getChromosome(int i)

Gets the i-th chromosome

Parameters:

i - the rule set we want to retrieve

Returns:

the selected rule set

onePointCrossover

public void onePointCrossover(int cr1,
                              int cr2)

One-point crossover

Parameters:

cr1 - index of parent 1 in poblation

cr2 - index of parent 2 in poblation

crossOver

public void crossOver()

It performs a one point crossover in the new poblation, using adjacent chromosomes as parents

elitism

public void elitism()

Copy the survivorsPercent proportion of the old poblation into the bottom half of the new one

mutate

public void mutate()

Applies mutation in the new poblation

selection

public void selection()

Applies a roulette wheel selection

tournament_selection

public void tournament_selection()

Applies a tournament selection, with tournament size of 2

evaluate

public void evaluate()

Its evaluate the NEW poblation, using a metric which summarizes the train CR and test CR

run

public void run()

It runs the Single-attribute Evolution Module (SEM) algorithm to obtain a rule set of ONE attribute