首页 > 分享 > 李航统计分析AdaBoost 提升书例子java代码

李航统计分析AdaBoost 提升书例子java代码

萌宠菠菠乐园
2024-09-14 00:03

Java代码

package xigua;

import java.util.Arrays;

import scala.Tuple2;

/**
*
*/

public class AdaBoostSquaredExample
{

private static int dataLength = 10;
private static int numIterations = 6;

   private static double[] SAMPLE = new double[]{
           5.56, 5.70, 5.91, 6.40, 6.80, 7.05, 8.90, 8.70, 9.00, 9.05
   };

   public static void main( String[] args )
   {
       Data[] datas = genData( );
       Arrays.sort( datas, ( s1, s2 ) -> s1.x - s2.x );

       for ( int i = 0; i < datas.length; i++ )
       {
           // System.out.println( data[i] );
       }

double[] bins = genBins( datas );

double[] ws = new double[datas.length];
Arrays.fill( ws, 1.0 / datas.length );

double[] alphas = new double[numIterations];
Arrays.fill( alphas, 0.0 );

BaseFunction[] fs = new BaseFunction[numIterations];

       for ( int i = 0; i < numIterations; i++ )
       {
           Tuple2<BaseFunction, Double> t = findBestFunction( datas, bins );
           fs[i] = t._1;

updateData( datas, t._1( ) );

       }
       // Test
       double totalError = 0.0;
       for ( int i = 0; i < datas.length; i++ )
       {
           double value = testData( datas[i], fs );
           System.out.println( SAMPLE[i] + "===" + value );
           double temp = SAMPLE[i] - value;
           totalError = totalError + temp * temp;
       }
       System.out.println( "TotalError == " + totalError );
       System.out.println( "Done" );
   }

   private static double testData( Data data, BaseFunction[] fs )
   {
       double value = 0.0;
       for ( int i = 0; i < fs.length; i++ )
       {
           if ( data.x <= fs[i].bin )
           {
               value = value + fs[i].leftMean;
           }
           else
           {
               value = value + fs[i].rightMean;
           }
       }

return value;
}

   private static void updateData( Data[] data, BaseFunction f )
   {
       double total = 0.0;
       for ( int i = 0; i < data.length; i++ )
       {
           if ( data[i].x <= f.bin )
           {
               data[i].y = data[i].y - f.leftMean;
           }
           else
           {
               data[i].y = data[i].y - f.rightMean;
           }
       }

}

   private static Tuple2<BaseFunction, Double> findBestFunction( Data[] data,
           double[] bins )
   {
       double minSq = Double.MAX_VALUE;
       BaseFunction best = null;
       for ( int i = 0; i < bins.length; i++ )
       {
           BaseFunction f = new BaseFunction( bins[i] );
           double leftTotal = 0.0;
           int leftCount = 0;
           double rightTotal = 0.0;
           int rightCount = 0;
           for ( int j = 0; j < data.length; j++ )
           {
               if ( data[j].x < bins[i] )
               {
                   leftTotal = leftTotal + data[j].y;
                   leftCount++;
               }
               else
               {
                   rightTotal = rightTotal + data[j].y;
                   rightCount++;
               }

           }
           double leftMean = leftTotal / leftCount;
           double rightMean = rightTotal / rightCount;

           double totalSq = 0.0;
           for ( int j = 0; j < data.length; j++ )
           {
               if ( data[j].x < bins[i] )
               {
                   double value = data[j].y - leftMean;
                   totalSq = totalSq + value * value;
               }
               else
               {
                   double value = data[j].y - rightMean;
                   totalSq = totalSq + value * value;
               }
           }

           if ( totalSq < minSq )
           {
               best = f;
               minSq = totalSq;
               best.leftMean = leftMean;
               best.rightMean = rightMean;
           }

       }
       return new Tuple2<BaseFunction, Double>( best, minSq );
   }

   private static double[] genBins( Data[] data )
   {
       if ( data.length == 0 )
       {
           return new double[0];
       }
       double[] bins = new double[data.length - 1];
       for ( int i = 1; i < data.length; i++ )
       {
           bins[i - 1] = ( data[i - 1].x + data[i].x ) / 2.0;
       }
       return bins;
   }

   private static Data[] genData( )
   {
       Data[] retValue = new Data[dataLength];
       for ( int i = 0; i < dataLength; i++ )
       {
           retValue[i] = new Data( );
           retValue[i].x = i + 1;
           retValue[i].y = SAMPLE[i];

       }
       return retValue;
   }

private static class Data
{

private int x;
private double y;

       @Override
       public String toString( )
       {
           return "X == " + x + "And Y == " + y;
       }

}

private static class BaseFunction
{

private double bin;
private double leftMean, rightMean;

       BaseFunction( )
       {
           this( -1 );
       }

       BaseFunction( double b )
       {
           this.bin = b;
       }

}
}