br.unicamp.ic.recod.gpsi.measures.gpsiClusterSilhouetteScore.java Source code

Introduction

Here is the source code for br.unicamp.ic.recod.gpsi.measures.gpsiClusterSilhouetteScore.java

Source

/*
 * To change this license header, choose License Headers in Project Properties.
 * To change this template file, choose Tools | Templates
 * and open the template in the editor.
 */
package br.unicamp.ic.recod.gpsi.measures;

import java.util.Arrays;
import org.apache.commons.math3.ml.distance.ManhattanDistance;

/**
 *
 * @author ra163128
 */
public class gpsiClusterSilhouetteScore implements gpsiSampleSeparationScore {

    @Override
    public double score(double[][][] samples) {

        ManhattanDistance distance = new ManhattanDistance();
        double score = 0.0;
        int m = 0, i, j, k;

        for (i = 0; i < samples.length; i++)
            m += samples[i].length;

        byte[] clusterAssignment = new byte[m];
        double[][] values = new double[m][];

        double[][] distances = new double[m - 1][];
        for (i = 0; i < m - 1; i++)
            distances[i] = new double[m - i - 1];

        int mIndex = 0;

        for (byte label = 0; label < samples.length; label++)
            for (i = 0; i < samples[label].length; i++) {
                values[mIndex] = samples[label][i];
                clusterAssignment[mIndex] = label;
                for (j = 0; j < mIndex; j++)
                    distances[mIndex - j - 1][j] = distance.compute(values[mIndex], values[mIndex - j - 1]);
                mIndex++;
            }

        double dissimilarity[] = new double[samples.length];
        double a, b;
        for (i = 0; i < m; i++) {
            Arrays.fill(dissimilarity, 0.0);
            for (j = 0; j < m; j++)
                if (i != j) {
                    k = Math.min(i, j);
                    dissimilarity[clusterAssignment[j]] += distances[k][Math.max(i, j) - 1 - k];
                }

            a = dissimilarity[clusterAssignment[i]] / (samples[clusterAssignment[i]].length - 1);
            b = Double.MAX_VALUE;
            for (j = 0; j < dissimilarity.length; j++)
                if (clusterAssignment[i] != j)
                    b = Math.min(b, dissimilarity[j] / samples[j].length);

            score += Math.max(a, b) == 0.0 ? -1.0 : (b - a) / Math.max(a, b);

        }

        return score / m;

    }

}