evaluatepredictions.java

本算法是实现基于KNN的基因遗传算法

/*
 * To change this template, choose Tools | Templates
 * and open the template in the editor.
 */

package gaknn;
import gaknn.evaluator.Evaluator;
import gaknn.predictor.Predictor;
import org.jgap.*;
import gaknn.similarity.AbstractSimilarity;

/**
 *
 * @author Niro
 */
public class EvaluatePredictions extends FitnessFunction{
    int m_ChromosomeLength;
    AbstractSimilarity m_Similarity;
    Predictor m_Predictor;
    Evaluator m_Evaluator;
    int m_K;
    boolean m_FindK = true;
    
    public EvaluatePredictions(AbstractSimilarity sim, Predictor pred, Evaluator eval){
        super();
        m_Similarity = sim;
        m_Predictor = pred;
        m_Evaluator = eval;
    }
    
    public void FindK(boolean find){
        m_FindK = find;
    }
    
    public void SetChromosomeLength(int length){
        m_ChromosomeLength = length;
    }
    
    public void SetSimilarity(AbstractSimilarity sim){
        m_Similarity = sim;
    }
    
    public void SetPredictor(Predictor pred){
        m_Predictor = pred;
    }
    
    public void SetK(int k){
        m_K = k;
    }
    
    public double evaluate(IChromosome a_chrom) {
        
        int weightVecSize;
        double fitness = 0.0;
        
        if (m_FindK){
            weightVecSize = a_chrom.size() - 1;
            m_K = (Integer) a_chrom.getGene(weightVecSize).getAllele();
        }
        else{
            weightVecSize = a_chrom.size();
        }
        
        double[] weights = new double[weightVecSize];
        
        for (int i=0; i< weightVecSize; i++)
            weights[i] = ( (Double) a_chrom.getGene( i ).getAllele() );
        
        m_Predictor.setK(m_K);
        m_Similarity.SetWeights(weights);
        m_Predictor.setSimilarity(m_Similarity);
        m_Evaluator.SetPredictor(m_Predictor);
        
        try {
            fitness =  m_Evaluator.Evaluate();
        }
        catch( Exception e ) {
            System.out.println(" "+e.toString() );
        }

        return (( 1.0 / fitness)*100.0) ;
    }

}