<P>
A data mining operator. Values in <I>TheTargetAttribute</I> are used as 
target function values to train the SVM on examples that are formed with <I>ThePredictingAttributes</I>. All <I>ThePredictingAttributes</I> must belong to <I>TheInputConcept</I>. 
<I>TheOutputAttribute</I> contains the predicted values.

<P>
There are some SVM-specific parameters; the table gives reasonable values to 
choose if nothing is known about the data or SVMs. For the <I>KernelType</I>, 
only the following values (Strings) are possible: <I>dot, polynomial, neural, 
radial, anova</I>. <I>Dot</I> is the linear kernel and can be taken as default. 

<P>
This operator can use two different versions of the Support Vector Machine algorithm.
One runs in main memory; it needs the parameter <I>SampleSize</I> to determine a maximum number of training examples.
The other runs in the database; it is used if the optional parameter <I>UseDB_SVM</I> is set to the String 
<TT>true</TT>. When this version is used, an additional parameter <I>TheKey</I> is needed which gives
the <TT>BaseAttribute</TT> whose column is the primary key of <I>TheInputConcept</I>. (<I>TheKey</I> can
be left out only if the <TT>ColumnSet</TT> that belongs to <I>TheInputConcept</I> represents a table rather than a view.)
The database algorithm restricts the possible kernel types to <I>dot</I> and <I>radial</I>. It can also use
the parameter <I>SampleSize</I>.

<P> With the parameters <I>LossFunctionPos</I> and <I>LossFunctionNeg</I>, the 
  loss function that is used for the regression can be biased such that predicting 
  too high is more expensive (<code>LossFunctionPos > LossFunctionNeg</code>) 
  or less expensive (<code>LossFunctionNeg > LossFunctionPos</code>) than predicting 
  too low. If both values are equal, no bias is used. The parameter <I>C</I> balances 
  training error against generalisation quality; positive values between 0.01 
  and 1000 have been used successfully in the literature. <I>Epsilon</I> limits 
  the allowed error an example may produce; small values under 0.5 should be used. 
<P>
<BR>
<BR>

<P>
PARAMETERDESCRIPTION
TheInputConcept:inherited
TheTargetAttribute:inherited
ThePredictingAttributes:&nbsp;
KernelType:see explanation above
SampleSize:see explanation above
LossFunctionPos:positive real; try 1.0
LossFunctionNeg:positive real; try 1.0
C:positive real; try 1.0
Epsilon:positive real; try 0.1
UseDB_SVM:
TheKey:optional
TheOutputAttribute:inherited
