clusteringinfosfuzzy.java

一个用于排队系统仿真的开源软件,有非常形象的图象仿真过程!

package jmt.engine.jwat.workloadAnalysis.clustering.fuzzyKMean;

import jmt.engine.jwat.MatrixOsservazioni;
import jmt.engine.jwat.Observation;
import jmt.engine.jwat.workloadAnalysis.clustering.ClusteringInfos;


public class ClusteringInfosFuzzy implements ClusteringInfos{

	public ClusterInfoFuzzy[] infoCluster;
	private double[][] assignment = null;
	private double err;
	private int numCluster;
	private short assignedToClust[];
	
	public int[] numElem;
	public double[]	percent;		// percentuale sul totale
	
	public String centri;					// contiene i centri dei cluster
	public String log;					// informazioni generali

	
	public ClusteringInfosFuzzy(double[][] ass,int nCluster){
		assignment = ass;
		numCluster = nCluster;
		err=-1;
		percent= new double[nCluster+1];
		assignedToClust=new short[assignment[0].length];
		infoCluster= new ClusterInfoFuzzy[nCluster];
	}
	
	public int getNumClusters(){
		return numCluster;
	}
	
	public ClusterInfoFuzzy getCluster(int index){
		if(err != -1 && index < numCluster){
			return infoCluster[index];
		}
		return null;
	}
	
	public short[] getAssignment()
	{
		return assignedToClust;
	}
	
	public void setError(MatrixOsservazioni m,double err){
		if(this.err==err) return;
		this.err = err;
		int i,j;
		boolean assigned;
		numElem = new int[numCluster+1];
		//Creazione InfoCluster
		for(i = 0; i < assignment[0].length;i++){
			assigned=false;
			for(j=0;j<assignment.length;j++){
	            if(assignment[j][i] >= err){
	            	 numElem[j]++;
	            	 if(!assigned) assignedToClust[i]=(short)(j+2);
	            	 else assignedToClust[i]=1;
	            	 assigned=true;
	            }
			}
			if(!assigned){
				assignedToClust[i]=0;
				numElem[numCluster]++;
			}
		}
		for(i=0;i<numCluster+1;i++){
			percent[i]=(double)numElem[i]/assignment[0].length;
		}
		InitInfo(m.getNumVariables());
		DoStat(m);
	}
	
	public double getError(){
		return err;
	}

	private void InitInfo(int nVar)
	{
		for(int i=0;i<infoCluster.length;i++){
			infoCluster[i]=new ClusterInfoFuzzy(nVar);
		}
	}
	private void DoStat(MatrixOsservazioni m){
		int nclus;
		
		Observation[] currOss=m.getListOss();		//Valore corrente di ogni variabile
		for (int count=0; count<m.getNumOfObs(); count++) {
			
			nclus = assignedToClust[count];
			if(nclus<2) continue;
			nclus-=2;
			infoCluster[nclus].numOss++;
			
			//Calcola somma, somma quadr. etc.
			
			for (int i=0; i<m.getNumVariables(); i++) {
				if (currOss[count].getIndex(i) !=0) {
					infoCluster[nclus].statClust[i].iNotZr++;
				}
				if (infoCluster[nclus].numOss==1) {
					infoCluster[nclus].statClust[i].dMaxOs = currOss[count].getIndex(i);
					infoCluster[nclus].statClust[i].dMinOs = currOss[count].getIndex(i);
				} else {
					if ( infoCluster[nclus].statClust[i].dMaxOs < currOss[count].getIndex(i) ) infoCluster[nclus].statClust[i].dMaxOs = currOss[count].getIndex(i);
					if ( infoCluster[nclus].statClust[i].dMinOs > currOss[count].getIndex(i) ) infoCluster[nclus].statClust[i].dMinOs = currOss[count].getIndex(i);
				}
				infoCluster[nclus].statClust[i].dSomma += currOss[count].getIndex(i);
				infoCluster[nclus].statClust[i].dSQuad += Math.pow(currOss[count].getIndex(i),2);
				infoCluster[nclus].statClust[i].dSTerz += Math.pow(currOss[count].getIndex(i),3);
				infoCluster[nclus].statClust[i].dSQuar += Math.pow(currOss[count].getIndex(i),4);
				
			}
		}
				
		double somma_per_var;
		//Calcola percentuale della variabile usata nel cluster
		for (int i=0; i < m.getNumVariables(); i++) {
			somma_per_var = 0;
			for (int l=0; l < numCluster; l++) {
				somma_per_var += infoCluster[l].statClust[i].dSomma;
			}
			
			for (int l=0; l < numCluster; l++) {
				
				infoCluster[l].percVar[i] = infoCluster[l].statClust[i].dSomma / somma_per_var;
			}
		}
		
		//Porting di stat2cl
		//Calcola media, varianza, etc,etc
		
		for (int l=0; l< numCluster; l++) {
			
			for (int i=0; i<m.getNumVariables(); i++) {
				infoCluster[l].statClust[i].dRange = infoCluster[l].statClust[i].dMaxOs - infoCluster[l].statClust[i].dMinOs;
				infoCluster[l].statClust[i].dMedia = infoCluster[l].statClust[i].dSomma / infoCluster[l].numOss;
				
				if (infoCluster[l].numOss != 1) {
					infoCluster[l].statClust[i].dPerc5 = Math.pow(infoCluster[l].statClust[i].dMedia,2);
					infoCluster[l].statClust[i].dPerc6 = Math.pow(infoCluster[l].statClust[i].dMedia,3);
					infoCluster[l].statClust[i].dPerc7 = Math.pow(infoCluster[l].statClust[i].dMedia,4);
					
					infoCluster[l].statClust[i].dVarnz = (infoCluster[l].statClust[i].dSQuad- infoCluster[l].numOss*infoCluster[l].statClust[i].dPerc5) / (infoCluster[l].numOss-1);
					
					if (infoCluster[l].statClust[i].dVarnz !=0) {
						infoCluster[l].statClust[i].dStdDv = Math.sqrt(infoCluster[l].statClust[i].dVarnz);
						infoCluster[l].statClust[i].dStdEr = infoCluster[l].statClust[i].dStdDv / Math.sqrt(infoCluster[l].numOss);
						
						infoCluster[l].statClust[i].dSkewn  = infoCluster[l].statClust[i].dSTerz - 3*infoCluster[l].statClust[i].dMedia*infoCluster[l].statClust[i].dSQuad + 3*infoCluster[l].statClust[i].dPerc5*infoCluster[l].statClust[i].dSomma;
						infoCluster[l].statClust[i].dSkewn /= infoCluster[l].numOss;
						infoCluster[l].statClust[i].dSkewn -= infoCluster[l].statClust[i].dPerc6;
						infoCluster[l].statClust[i].dSkewn /= Math.pow(infoCluster[l].statClust[i].dVarnz,1.5);
						
						infoCluster[l].statClust[i].dKurto  = infoCluster[l].statClust[i].dSQuar - 4*infoCluster[l].statClust[i].dMedia*infoCluster[l].statClust[i].dSTerz + 6*infoCluster[l].statClust[i].dPerc5*infoCluster[l].statClust[i].dSQuad - 4*infoCluster[l].statClust[i].dPerc6*infoCluster[l].statClust[i].dSomma;
						infoCluster[l].statClust[i].dKurto /= infoCluster[l].numOss;
						infoCluster[l].statClust[i].dKurto += infoCluster[l].statClust[i].dPerc7;
						infoCluster[l].statClust[i].dKurto /= Math.pow(infoCluster[l].statClust[i].dVarnz,2);
						infoCluster[l].statClust[i].dKurto -= 3;
					}
				}
			}
		}
	}

}