attributevisualizationpanel.java

一个数据挖掘软件ALPHAMINERR的整个过程的JAVA版源代码

	      maxValue = as.nominalCounts[i];
	  }
	  
	  for(int k=0; k<m_data.numInstances(); k++) {
	  	if(!m_data.instance(k).isMissing(attribIndex))
	  	{
	  		histCounts[(int)m_data.instance(k).value(attribIndex)]++;
	  	}
	  }
	  threadRun=false;
	  histBarCounts = histCounts;
	  //Image tmpImg = new BufferedImage(getWidth(), getHeight(), BufferedImage.TYPE_INT_RGB);
	  //drawGraph( tmpImg.getGraphics() );
	  //img = tmpImg;
	  AttributeVisualizationPanel.this.repaint();
	}
      }
    }
  }


  private class HistCalc extends Thread {

   /****Code for M.P.Wand's method of histogram bin width selection.
    *   There is some problem with it. It always comes up -ve value 
    *   which is raised to the power 1/3 and gives an NAN.
    private static final int M=400;
    private double psi(int r, double g) {
      double val;
      
      double sum=0.0;
      for(int i=0; i<M; i++) {
	  double valCjKj=0.0;
	  for(int j=0; j<M; j++) {
	      valCjKj += c(j) * k(r, j-i, g);
	  }
	  sum += valCjKj*c(i);
      }

      val = Math.pow(m_data.numInstances(), -2) * sum;
      //System.out.println("psi returns: "+val);
      return val;
    }
    private double g21() {
      double val;
      
      val = Math.pow(2 / ( Math.sqrt(2D*Math.PI)*psi(4, g22())*m_data.numInstances() ), 1/5D) * Math.sqrt(2) * as.numericStats.stdDev;
      //System.out.println("g21 returns: "+val);
      return val;
    }    
    private double g22() {
      double val;

      val = Math.pow( 2D/(5*m_data.numInstances()), 1/7D) * Math.sqrt(2) * as.numericStats.stdDev;
      //System.out.println("g22 returns: "+val);
      return val;
    }
    private double c(int j) {
      double val=0.0;
      double sigma = (as.numericStats.max - as.numericStats.min)/(M-1);
      
      //System.out.println("In c before doing the sum we have");
      //System.out.println("max: " +as.numericStats.max+" min: "+as.numericStats.min+
      //		 " sigma: "+sigma);
      for(int i=0; i<m_data.numInstances(); i++) {
	  if(!m_data.instance(i).isMissing(attribIndex))
	      val += Math.max( 0, 
			       ( 1 - Math.abs( Math.pow(sigma, -1)*(m_data.instance(i).value(attribIndex) - j) ) ) 
			     );
      }
      //System.out.println("c returns: "+val);
      return val;
    }
    private double k(int r, int j, double g) {
      double val;
      double sigma = (as.numericStats.max - as.numericStats.min)/(M-1);
      //System.out.println("Before calling L we have");
      //System.out.println("Max: "+as.numericStats.max+" Min: "+as.numericStats.min+"\n"+
      //			 "r: "+r+" j: "+j+" g: "+g);
      val = Math.pow( g, -r-1) * L(sigma*j/g);
      //System.out.println("k returns: "+val);
      return val;	
    }
    private double L(double x) {
      double val;
      
      val = Math.pow( 2*Math.PI, -1/2D ) * Math.exp( -(x*x)/2D );
      //System.out.println("L returns: "+val);
      return val;
    }
    *******End of Wand's method
    */

    public void run() {
      synchronized (m_locker) {
	if((classIndex >= 0) && (m_data.attribute(classIndex).isNominal())) {
	     
	  int intervals; double intervalWidth=0.0;

	  //This uses the M.P.Wand's method to calculate the histogram's interval width.
	  //See "Data-Based Choice of Histogram Bin Width".
	  //intervalWidth = Math.pow( 6D/( -psi(2, g21()) * m_data.numInstances() ), 1/3D );

	  //This uses the Scott's method to calculate the histogram's interval width.
	  //See "On optimal and data-based histograms". Biometrika, 66, 605-610 or 
	  //see the same paper mentioned above.	  
	  intervalWidth =  3.49 * as.numericStats.stdDev * Math.pow(m_data.numInstances(), -1/3D);
	  //The Math.max is introduced to remove the possibility of intervals=0 that can happen
	  //if the intervalWidth<1
	  intervals = Math.max(1, (int)Math.round((as.numericStats.max - as.numericStats.min)/intervalWidth));
	    
	  //System.out.println("Max: "+as.numericStats.max+" Min: "+as.numericStats.min+" stdDev: "+as.numericStats.stdDev+
	  //	     "intervalWidth: "+intervalWidth);
	  if(intervals > AttributeVisualizationPanel.this.getWidth()) {
	      intervals = AttributeVisualizationPanel.this.getWidth()-4;
	      if(intervals<1)
		  intervals = 1;
	  }
	  int histClassCounts[][]  = new int[intervals][m_data.attribute(classIndex).numValues()+1];
	  double barRange   = (as.numericStats.max - as.numericStats.min)/(double)histClassCounts.length;
	  double currentBar = as.numericStats.min; // + barRange;
	  maxValue = 0;
  
	  if(m_colorList.size()==0)
	    m_colorList.addElement(Color.black);
	  for(int i=m_colorList.size(); i<m_data.attribute(classIndex).numValues()+1; i++) {
	    Color pc = m_defaultColors[(i-1) % 10];
	    int ija =  (i-1) / 10;
	    ija *= 2; 
	    for (int j=0;j<ija;j++) {
	      pc = pc.darker();     
	    }
	    m_colorList.addElement(pc);
	  }
	  
	  for(int k=0; k<m_data.numInstances(); k++) 
	  {
	    int t=0;
	    try {
	    	if(!m_data.instance(k).isMissing(attribIndex)) {
	      	t = (int) Math.ceil((float)((m_data.instance(k).value(attribIndex)-as.numericStats.min)/barRange));
	      	if(t==0) 
	      	{
	      		if(m_data.instance(k).isMissing(classIndex))
	      			histClassCounts[t][0]++;
	      		else
	      			histClassCounts[t][(int)m_data.instance(k).value(classIndex)+1]++;
			  //if(histCounts[t]>maxValue)
			  //  maxValue = histCounts[t];
	      	} else 
	      	{
	      		if(m_data.instance(k).isMissing(classIndex))
	      			histClassCounts[t-1][0]++;
	      		else
	      			histClassCounts[t-1][(int)m_data.instance(k).value(classIndex)+1]++;
	      		//if(histCounts[t-1]>maxValue)
	      		//  maxValue = histCounts[t-1];
	      	}
	      }
	    }
	    catch(ArrayIndexOutOfBoundsException ae) { 
	    	System.out.println("t:"+(t)+
	    			 " barRange:"+barRange+
	    			 " histLength:"+histClassCounts.length+
	    			 " value:"+m_data.instance(k).value(attribIndex)+
	    			 " min:"+as.numericStats.min+
	    			 " sumResult:"+(m_data.instance(k).value(attribIndex)-as.numericStats.min)+
	    			 " divideResult:"+(float)((m_data.instance(k).value(attribIndex)-as.numericStats.min)/barRange)+
	    			 " finalResult:"+
	    			 Math.ceil((float)((m_data.instance(k).value(attribIndex)-as.numericStats.min)/barRange)) ); }
	  }
	  for(int i=0; i<histClassCounts.length; i++) {
	    int sum=0;
	    for(int j=0; j<histClassCounts[i].length; j++) 
	      sum = sum+histClassCounts[i][j];
	    if(maxValue<sum)
	      maxValue = sum;
	  }
	  histBarClassCounts = histClassCounts;
	  m_barRange =  barRange;
	  
	}
	else { //else if the class attribute is numeric or the class is not set

	  int intervals =  as.totalCount>10 ? 
	                  (int)(as.totalCount*0.1):(int)as.totalCount;//At the time of this coding the
	                                       //possibility of datasets with zero instances 
	                                       //was being dealt with in the 
	                                       //PreProcessPanel of weka Explorer.

	  if(intervals > AttributeVisualizationPanel.this.getWidth()) {
	      intervals = AttributeVisualizationPanel.this.getWidth()-4;
	      if(intervals<1)
		  intervals = 1;
	  }

	  int histCounts[]  = new int[intervals];
	  double barRange   = (as.numericStats.max - as.numericStats.min)/(double)histCounts.length;
	  double currentBar = as.numericStats.min; // + barRange;
	  maxValue = 0;
	  
	  for(int k=0; k<m_data.numInstances(); k++) {
	    int t=0;
	    try {
/*	    	if (Double.isNaN(m_data.instance(k).value(attribIndex)))
	    	{
	    		// Skip NaN
	    		continue;
	    	}*/
	    	if(!m_data.instance(k).isMissing(attribIndex)) 
	    	{
	    		t = (int) Math.ceil((float)((m_data.instance(k).value(attribIndex)-as.numericStats.min)/barRange));
	    		if(t==0) 
	    		{
					histCounts[t]++;
					if(histCounts[t]>maxValue)
					  maxValue = histCounts[t];
				} else {
					histCounts[t-1]++;
					if(histCounts[t-1]>maxValue)
						maxValue = histCounts[t-1];
				}
		    }
	    } catch(ArrayIndexOutOfBoundsException ae) { 
		      ae.printStackTrace();
		      System.out.println("t:"+(t)+
					 " barRange:"+barRange+
					 " histLength:"+histCounts.length+
					 " value:"+m_data.instance(k).value(attribIndex)+
					 " min:"+as.numericStats.min+
					 " sumResult:"+(m_data.instance(k).value(attribIndex)-as.numericStats.min)+
					 " divideResult:"+(float)((m_data.instance(k).value(attribIndex)-as.numericStats.min)/barRange)+
					 " finalResult:"+
					 Math.ceil((float)((m_data.instance(k).value(attribIndex)-as.numericStats.min)/barRange)) ); }
	    
	  	}
	  	histBarCounts = histCounts;
	  	m_barRange =  barRange;
	}
	
	threadRun=false;
	//Image tmpImg = new BufferedImage(getWidth(), getHeight(), BufferedImage.TYPE_INT_RGB);
	//drawGraph( tmpImg.getGraphics() );
	//img = tmpImg;
	AttributeVisualizationPanel.this.repaint();
      }
    }
  }


  /**
   * Returns "&lt;nominal value&gt; [&lt;nominal value count&gt;]"
   * if displaying a bar plot and mouse is on some bar.
   * If displaying histogram then it 
   *     <li>returns "count &lt;br&gt; [&lt;bars Range&gt;]" if mouse is 
   *     on the first bar. </li>
   *     <li>returns "count &lt;br&gt; (&lt;bar's Range&gt;]" if mouse is 
   *     on some bar other than the first one. </li>
   * Otherwise it returns ""
   *
   * @param ev The mouse event
   */
  public String getToolTipText(MouseEvent ev) {
    if(as!=null && as.nominalCounts!=null) {
      float intervalWidth = this.getWidth()/(float)as.nominalCounts.length, heightRatio;
      int barWidth, x=0, y=0;
      if(intervalWidth>5)
	barWidth = (int)Math.floor(intervalWidth*0.8F);
      else
	barWidth = 1;

      x = x + (int)( (Math.floor(intervalWidth*0.1F))<1 ? 1:(Math.floor(intervalWidth*0.1F)) );
      if(  this.getWidth() - 
	   (x + as.nominalCounts.length*barWidth 
	    +(int)( (Math.floor(intervalWidth*0.2F))<1 ? 1:(Math.floor(intervalWidth*0.2F)) )*as.nominalCounts.length) > 5 )
	x += (this.getWidth() - 
	      (x + as.nominalCounts.length*barWidth + 
	       (int)( (Math.floor(intervalWidth*0.2F))<1 ? 1:(Math.floor(intervalWidth*0.2F)) )*as.nominalCounts.length))/2;
      for(int i=0; i<as.nominalCounts.length; i++) {
	heightRatio = (this.getHeight()-(float)fm.getHeight())/maxValue;
	y = this.getHeight()-Math.round(as.nominalCounts[i]*heightRatio);
	if(ev.getX()>=x && ev.getX()<=x+barWidth
	   && ev.getY()>=this.getHeight()-Math.round(as.nominalCounts[i]*heightRatio) )
	  return(m_data.attribute(attribIndex).value(i)+" ["+as.nominalCounts[i]+"]");
	x = x+barWidth+(int)( (Math.floor(intervalWidth*0.2F))<1 ? 1:(Math.floor(intervalWidth*0.2F)) );
      }
    }
    else if(threadRun==false && (histBarCounts!=null || histBarClassCounts!=null)) {
      float heightRatio, intervalWidth;
      int x=0, y=0,  barWidth;
      double bar = as.numericStats.min;
      
      if((classIndex >= 0) && (m_data.attribute(classIndex).isNominal())) {
	barWidth = ((this.getWidth()-6)/histBarClassCounts.length)<1 ? 1:((this.getWidth()-6)/histBarClassCounts.length);
	  
	x = 3;
	if( (this.getWidth() - (x + histBarClassCounts.length*barWidth)) > 5 )
	  x += (this.getWidth() - (x + histBarClassCounts.length*barWidth))/2;
	  
	heightRatio = (this.getHeight()-(float)fm.getHeight())/maxValue;
	  
	if( ev.getX()-x >= 0) {
	  int temp = (int)((ev.getX()-x)/(barWidth+0.0000000001));
	  if(temp == 0){