package jsat.regression.evaluation;
import jsat.math.OnLineStatistics;
/**
* Uses the Mean of Absolute Errors between the predictions and the true values.
*
* @author Edward Raff
*/
public class MeanAbsoluteError implements RegressionScore
{
private static final long serialVersionUID = -637676526509989776L;
private OnLineStatistics absError;
public MeanAbsoluteError()
{
}
/**
* Copy constructor
* @param toCopy the object to copy
*/
public MeanAbsoluteError(MeanAbsoluteError toCopy)
{
if(toCopy.absError != null)
this.absError = toCopy.absError.clone();
}
@Override
public void prepare()
{
absError = new OnLineStatistics();
}
@Override
public void addResult(double prediction, double trueValue, double weight)
{
if(absError == null)
throw new RuntimeException("regression score has not been initialized");
absError.add(Math.abs(prediction-trueValue), weight);
}
@Override
public void addResults(RegressionScore other)
{
MeanAbsoluteError otherObj = (MeanAbsoluteError) other;
if(otherObj.absError != null)
this.absError.add(otherObj.absError);
}
@Override
public double getScore()
{
return absError.getMean();
}
@Override
public boolean lowerIsBetter()
{
return true;
}
@Override
public MeanAbsoluteError clone()
{
return new MeanAbsoluteError(this);
}
@Override
public int hashCode()
{//XXX this is a strange hashcode method
return getName().hashCode();
}
@Override
public boolean equals(Object obj)
{//XXX check for equality of fields and obj == null
if(this.getClass().isAssignableFrom(obj.getClass()) && obj.getClass().isAssignableFrom(this.getClass()))
{
return true;
}
return false;
}
@Override
public String getName()
{
return "Mean Absolute Error";
}
}
Ads help maintain this website.