package com.rapidminer.operator.ansamble;
import java.util.ArrayList;
import java.util.Iterator;
import java.util.List;
import com.rapidminer.operator.Operator;
import com.rapidminer.operator.OperatorDescription;
import com.rapidminer.operator.OperatorException;
import com.rapidminer.operator.RatingPrediction.GroupRecommender;
import com.rapidminer.operator.RatingPrediction.RatingPredictor;
import com.rapidminer.operator.ports.InputPort;
import com.rapidminer.operator.ports.InputPortExtender;
import com.rapidminer.operator.ports.OutputPort;
import com.rapidminer.operator.ports.metadata.GenerateNewMDRule;
import com.rapidminer.operator.ports.metadata.MetaData;
import com.rapidminer.operator.ports.metadata.Precondition;
import com.rapidminer.operator.ports.metadata.SimplePrecondition;
import com.rapidminer.parameter.ParameterType;
import com.rapidminer.parameter.ParameterTypeDouble;
import com.rapidminer.parameter.ParameterTypeList;
import com.rapidminer.parameter.ParameterTypeString;
/**
* ModelCombiner operator for RatingPrediction
*
* @see com.rapidminer.operator.ansamble.ModelCombiner
*
* @author Matej Mihelcic (Ru�er Bo�kovi� Institute)
*/
public class ModelCombiner extends Operator{
public static final String PARAMETER_DEFAULT_WEIGHT = "default_weight";
public static final String PARAMETER_MODEL_WEIGHTS = "model_weights";
private final InputPortExtender inputPort=new InputPortExtender("model",getInputPorts()){
@Override
protected Precondition makePrecondition(InputPort port) {
int index = inputPort.getManagedPorts().size();
return new SimplePrecondition(port, new MetaData(RatingPredictor.class), index < 2);
};
};
private OutputPort exampleSetOutput = getOutputPorts().createPort("grouped model");
public ModelCombiner(OperatorDescription description) {
super(description);
inputPort.start();
inputPort.ensureMinimumNumberOfPorts(2);
MetaData met=new MetaData(RatingPredictor.class);
inputPort.getManagedPorts().get(0).addPrecondition(new SimplePrecondition(inputPort.getManagedPorts().get(0), met));
inputPort.getManagedPorts().get(1).addPrecondition(new SimplePrecondition(inputPort.getManagedPorts().get(1), met));
getTransformer().addRule(new GenerateNewMDRule(exampleSetOutput, new MetaData(RatingPredictor.class)) {
});
}
@Override
public List<ParameterType> getParameterTypes() {
List<ParameterType> types = super.getParameterTypes();
types.add(new ParameterTypeDouble(PARAMETER_DEFAULT_WEIGHT, "The default weight for all models not specified in the list 'model_weights'.", 0.0d, Double.POSITIVE_INFINITY, 1.0d));
types.add(new ParameterTypeList(PARAMETER_MODEL_WEIGHTS, "The weights for several models. Criteria weights not defined in this list are set to 'default_weight'.",
new ParameterTypeString("operator_name", "The name of the operator."),
new ParameterTypeDouble("model_weight", "The weight for this model.", 0.0d,
Double.POSITIVE_INFINITY, 1.0d)));
return types;
}
@Override
public void doWork() throws OperatorException{
List<RatingPredictor> ansambl=inputPort.getData(true);
List<Double> weights=new ArrayList<Double>();
List<String[]> weightList = getParameterList(PARAMETER_MODEL_WEIGHTS);
Iterator<String[]> i = weightList.iterator();
while (i.hasNext()) {
String[] entry = i.next();
Double criterionWeight = Double.valueOf(entry[1]);
weights.add(criterionWeight);
}
GroupRecommender recommendAlg=new GroupRecommender();
recommendAlg.SetRatings(ansambl.get(0).GetRatings());
recommendAlg.SetMaxRating(ansambl.get(0).GetMaxRating());
recommendAlg.SetMinRating(ansambl.get(0).GetMinRating());
recommendAlg.SetRecommenders(ansambl);
recommendAlg.item_mapping=ansambl.get(0).item_mapping;
recommendAlg.user_mapping=ansambl.get(0).user_mapping;
recommendAlg.SetWeights(weights);
recommendAlg.SetDWeight(getParameterAsDouble("default_weight"));
exampleSetOutput.deliver(recommendAlg);
}
}