EM Algorithm

/*
 * Kocaeli University Computer Engineering Department 2011-Fall
 *
 * Class     : Data Mining
 * Subject   : Inspection and Implementation of Model Based Clustering Method
 *             Expectation-Maximization (EM)
 *
 * Student   : Mustafa KIYAR
 * Number    : 105112014
 *
 */
package em;

import java.io.BufferedWriter;
import java.io.FileInputStream;
import java.io.FileWriter;
import java.io.IOException;
import java.util.Properties;
import org.apache.commons.lang3.ArrayUtils;

/**
 *
 * @author Mustafa KIYAR
 */
public class EM {

    public static double findProbablity(double x, double mu, double sigma) {
        double dExp = Math.exp(-Math.pow(x - mu, 2) / (2 * Math.pow(sigma, 2)));
        return dExp / (Math.sqrt(2 * Math.PI) * sigma);
    }

    public static double findMean(double dArr[]) {
        double mu = 0;
        for (int i = 0; i < dArr.length; i++) {
            mu += dArr[i];
        }
        return mu / dArr.length;
    }

    public static double findStandartDeviation(double dArr[], double mu) {
        double sigma = 0;
        for (int i = 0; i < dArr.length; i++) {
            sigma += Math.pow(dArr[i] - mu, 2);
        }
        return Math.sqrt(sigma / dArr.length);
    }

    public static void myLog(String str) {
        System.out.println(str);
    }

    public static void analyzeArr(double dArr[]) {
        double mu, sigma;

        mu = findMean(dArr);
        sigma = findStandartDeviation(dArr, mu);

        myLog("mean               : " + mu);
        myLog("standart deviation : " + sigma);

        for (int i = 0; i < dArr.length; i++) {
            myLog("probablity of " + dArr[i] + " is "
                    + findProbablity(dArr[i], mu, sigma));
        }
    }

    public static double findProbablity(double x, double dArr[]) {
        double mu, sigma, probablity;

        mu = findMean(dArr);
        sigma = findStandartDeviation(dArr, mu);
        probablity = findProbablity(x, mu, sigma);

        myLog("mean                 : " + mu);
        myLog("standart deviation   : " + sigma);
        myLog("probablity of object : " + probablity);
        return probablity;
    }

    public static int findRelation(double x, double dArr[][]) {
        double nominator = 0, denominator = 0, probablity = 0;
        int iBestCluster = 0;
        for (int i = 0; i < dArr.length; i++) {
            myLog("");
            myLog("Inspection of " + x + " in Cluster :" + i);
            probablity = findProbablity(x, dArr[i]);
            denominator += probablity;
            if (nominator < probablity) {
                nominator = probablity;
                iBestCluster = i;
            }
        }
        return iBestCluster;
    }

    public static void switchCluster(int iCluster, int iPos, int iBestCluster,
            double dArr[][]) {

        double dVal = dArr[iCluster][iPos];
        dArr[iCluster] = ArrayUtils.remove(dArr[iCluster], iPos);
        dArr[iBestCluster] = ArrayUtils.add(dArr[iBestCluster], dVal);

        myLog("Object           : " + dVal);
        myLog("Original cluster : " + iCluster);
        myLog("Moved to cluster : " + iBestCluster);

    }

    public static void runProgram(double dArr[][]) {
        int iBestCluster;
        boolean check;
        do {
            check = false;
            for (int i = 0; i < dArr.length; i++) {
                myLog("##########################################################");
                myLog("Scaning Cluster : " + i);
                for (int j = 0; j < dArr[i].length; j++) {
                    iBestCluster = findRelation(dArr[i][j], dArr);
                    if (i != iBestCluster) {
                        switchCluster(i, j, iBestCluster, dArr);
                        j--;
                        check = true;
                    } else {
                        myLog("Object belongs to orginal cluster.");
                    }
                    myLog("......................................................");
                }
            }
        } while (check);
    }

    /**
     * @param args the command line arguments
     */
    public static void main(String[] args) {

        String sArrInput[] = {""};
        int numberOfCluster = 0, iSizeCuster = 0;
        Properties property = new Properties();
        try {
            property.load(new FileInputStream(args[0]));
            sArrInput = property.getProperty("data").split(",");
            numberOfCluster = Integer.parseInt(property.getProperty("numberofclusters"));

        } catch (IOException ex) {
            ex.printStackTrace();
        }

        if (sArrInput.length < 3 || sArrInput.length > 1000
                || sArrInput.length <= (numberOfCluster * 2)) {
            myLog("Input data is irrelevant!");
            return;
        }

        /* find size of an cluster */
        iSizeCuster = sArrInput.length / numberOfCluster;

        double dArr[][] = new double[numberOfCluster][iSizeCuster];
        double dCluster[] = new double[iSizeCuster];
        for (int i = 0; i < sArrInput.length; i++) {
            dCluster[i % iSizeCuster] = Double.parseDouble(sArrInput[i]);
            if (i != 0 && (i + 1) % iSizeCuster == 0) {
                dArr[i / iSizeCuster] = dCluster;
                dCluster = new double[iSizeCuster];
            }
        }
        if (sArrInput.length % iSizeCuster != 0) {
            dCluster = ArrayUtils.subarray(dCluster, 0, (sArrInput.length % iSizeCuster));
            dArr[numberOfCluster - 1] = ArrayUtils.addAll(dArr[numberOfCluster - 1], dCluster);
        }

        runProgram(dArr);
        try {

            FileWriter fw = new FileWriter("out.log");
            BufferedWriter bw = new BufferedWriter(fw);

            double dStdDev = 0.0, dMean = 0.0;

            for (int i = 0; i < dArr.length; i++) {
                dMean = findMean(dArr[i]);
                dStdDev = findStandartDeviation(dArr[i], dMean);
                bw.write("Cluster " + i + " : " + ArrayUtils.toString(dArr[i]) + "\n");
                bw.write("Mean of Cluster " + i + " : " + Double.toString(dMean)+ "\n");
                bw.write("Stdev of Cluster " + i + " : " + Double.toString(dStdDev)+"\n");
                bw.write("\n");
            }
            bw.close();

        } catch (IOException ex) {
            ex.printStackTrace();
        }
    }
}