/*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*/
/*
* Meta.java
* Copyright (C) 2009 Aristotle University of Thessaloniki, Thessaloniki, Greece
*/
package mulan.classifier.meta.thresholding;
import java.util.ArrayList;
import java.util.Collections;
import java.util.logging.Level;
import java.util.logging.Logger;
import mulan.classifier.InvalidDataException;
import mulan.classifier.ModelInitializationException;
import mulan.classifier.meta.*;
import mulan.classifier.MultiLabelLearner;
import mulan.classifier.MultiLabelOutput;
import mulan.data.DataUtils;
import mulan.data.MultiLabelInstances;
import mulan.transformations.RemoveAllLabels;
import weka.classifiers.Classifier;
import weka.core.Attribute;
import weka.core.Instance;
import weka.core.Instances;
/**
* Base class for instance-based prediction of a bipartition from
* the labels' scores
*
* @author Marios Ioannou
* @author George Sakkas
* @author Grigorios Tsoumakas
* @version 2010.12.14
*/
public abstract class Meta extends MultiLabelMetaLearner {
/** the classifier to learn the number of top labels or the threshold */
protected Classifier classifier;
/** the training instances for the single-label model */
protected Instances classifierInstances;
/** the type for constructing the meta dataset*/
protected String metaDatasetChoice;
/**the number of folds for cross validation*/
protected int kFoldsCV;
/** clean multi-label learner for cross-validation */
protected MultiLabelLearner foldLearner;
/**
* Constructor that initializes the learner
*
* @param baseLearner the MultiLabelLearner
* @param aClassifier the learner that will predict the number of relevant
* labels or a threshold
* @param aMetaDatasetChoice what features to use for predicting the number
* of relevant labels or a threshold
*/
public Meta(MultiLabelLearner baseLearner, Classifier aClassifier, String aMetaDatasetChoice) {
super(baseLearner);
metaDatasetChoice = aMetaDatasetChoice;
classifier = aClassifier;
}
/**
* Returns the classifier used to predict the number of labels/threshold
*
* @return the classifier used to predict the number of labels/threshold
*/
public Classifier getClassifier() {
return classifier;
}
/**
* abstract method that transforms the training data to meta data
*
* @param trainingData the training data set
* @return the meta data for training the predictor of labels/threshold
* @throws Exception
*/
protected abstract Instances transformData(MultiLabelInstances trainingData) throws Exception;
/**
* A method that modify an instance
*
* @param instance to modified
* @param xBased the type for constructing the meta dataset
* @return a transformed instance for the predictor of labels/threshold
*/
protected Instance modifiedInstanceX(Instance instance, String xBased) {
Instance modifiedIns = null;
MultiLabelOutput mlo = null;
if (xBased.compareTo("Content-Based") == 0) {
Instance tempInstance = RemoveAllLabels.transformInstance(instance, labelIndices);
modifiedIns = DataUtils.createInstance(tempInstance, tempInstance.weight(), tempInstance.toDoubleArray());
} else if (xBased.compareTo("Score-Based") == 0) {
double[] arrayOfScores = new double[numLabels];
try {
mlo = baseLearner.makePrediction(instance);
} catch (InvalidDataException ex) {
Logger.getLogger(Meta.class.getName()).log(Level.SEVERE, null, ex);
} catch (ModelInitializationException ex) {
Logger.getLogger(Meta.class.getName()).log(Level.SEVERE, null, ex);
} catch (Exception ex) {
Logger.getLogger(Meta.class.getName()).log(Level.SEVERE, null, ex);
}
arrayOfScores = mlo.getConfidences();
modifiedIns = DataUtils.createInstance(instance, numLabels);
for (int i = 0; i < numLabels; i++) {
modifiedIns.setValue(i, arrayOfScores[i]);
}
} else { //Rank-Based
try {
//Rank-Based
double[] arrayOfScores = new double[numLabels];
mlo = baseLearner.makePrediction(instance);
arrayOfScores = mlo.getConfidences();
ArrayList<Double> list = new ArrayList();
for (int i = 0; i < numLabels; i++) {
list.add(arrayOfScores[i]);
}
Collections.sort(list);
modifiedIns = DataUtils.createInstance(instance, numLabels);
int j = numLabels - 1;
for (Double x : list) {
modifiedIns.setValue(j, x);
j--;
}
} catch (InvalidDataException ex) {
Logger.getLogger(Meta.class.getName()).log(Level.SEVERE, null, ex);
} catch (ModelInitializationException ex) {
Logger.getLogger(Meta.class.getName()).log(Level.SEVERE, null, ex);
} catch (Exception ex) {
Logger.getLogger(Meta.class.getName()).log(Level.SEVERE, null, ex);
}
}
return modifiedIns;
}
/**
* Prepares the instances for the predictor of labels/threshold
*
* @param data the training data
* @return the prepared instances
*/
protected Instances prepareClassifierInstances(MultiLabelInstances data) {
Instances temp = null;
if (metaDatasetChoice.compareTo("Content-Based") == 0) {
try {
temp = RemoveAllLabels.transformInstances(data);
temp = new Instances(temp, 0);
} catch (Exception ex) {
Logger.getLogger(Meta.class.getName()).log(Level.SEVERE, null, ex);
}
} else {
ArrayList<Attribute> atts = new ArrayList<Attribute>();
for (int i=0; i<numLabels; i++) {
atts.add(new Attribute("Label" + i));
}
temp = new Instances("threshold", atts, 0);
}
return temp;
}
/**
* A method that fill the array "newValues"
*
* @param learner the multi-label learner
* @param instance the training instances
* @param newValues the array to fill
* @param xBased the type for constructing the meta dataset
* @throws Exception
*/
protected void valuesX(MultiLabelLearner learner, Instance instance, double[] newValues, String xBased) throws Exception {
MultiLabelOutput mlo = null;
if (metaDatasetChoice.compareTo("Content-Based") == 0) {
double[] values = instance.toDoubleArray();
for (int i=0; i<featureIndices.length; i++)
newValues[i] = values[featureIndices[i]];
} else if (metaDatasetChoice.compareTo("Score-Based") == 0) {
mlo = learner.makePrediction(instance);
double[] values = mlo.getConfidences();
System.arraycopy(values, 0, newValues, 0, values.length);
} else if (metaDatasetChoice.compareTo("Rank-Based") == 0) {
mlo = learner.makePrediction(instance);
double[] values = mlo.getConfidences();
ArrayList<Double> list = new ArrayList();
for (int i = 0; i < numLabels; i++) {
list.add(values[i]);
}
Collections.sort(list);
int j = numLabels - 1;
for (Double x : list) {
newValues[j] = x;
j--;
}
}
}
@Override
protected void buildInternal(MultiLabelInstances trainingData) throws Exception {
// build the base multilabel learner from the original training data
baseLearner.build(trainingData);
classifierInstances = transformData(trainingData);
// build the prediction model
classifier.buildClassifier(classifierInstances);
// keep just the header information
classifierInstances = new Instances(classifierInstances, 0);
}
}