List of usage examples for weka.core Instance toDoubleArray
public double[] toDoubleArray();
From source file:mulan.classifier.meta.HMC.java
License:Open Source License
private void makePrediction(HMCNode currentNode, Instance instance, boolean[] predictedLabels, double[] confidences) throws Exception { //System.out.println("Node: " + currentNode.getName()); double[] values = instance.toDoubleArray(); Instance transformed = DataUtils.createInstance(instance, 1, values); // delete all labels apart from those of current node int[] currentNodeLabelIndices = currentNode.getLabelIndices(); Set<Integer> indicesToKeep = new HashSet<Integer>(); for (int i = 0; i < currentNodeLabelIndices.length; i++) { String labelToKeep = currentNode.getHeader().attribute(currentNodeLabelIndices[i]).name(); indicesToKeep.add(labelIndices[labelsAndIndices.get(labelToKeep)]); }//from ww w. ja va 2s .co m if (labelIndices.length - indicesToKeep.size() != 0) { int[] indicesToDelete = new int[labelIndices.length - indicesToKeep.size()]; int counter = 0; for (int i = 0; i < labelIndices.length; i++) { if (indicesToKeep.contains(labelIndices[i])) { continue; } indicesToDelete[counter] = labelIndices[i]; counter++; } transformed = RemoveAllLabels.transformInstance(transformed, indicesToDelete); } transformed.setDataset(currentNode.getHeader()); // add as many attributes as the children // System.out.println("header:" + currentNode.getHeader()); //System.out.println(transformed.toString()); //debug("working at node " + currentNode.getName()); //debug(Arrays.toString(predictedLabels)); NoClassifierEvals++; MultiLabelOutput pred = currentNode.makePrediction(transformed); int[] indices = currentNode.getLabelIndices(); boolean[] temp = pred.getBipartition(); for (int i = 0; i < temp.length; i++) { String childName = currentNode.getHeader().attribute(indices[i]).name(); //System.out.println("childName:" + childName); int idx = labelsAndIndices.get(childName); if (pred.getBipartition()[i] == true) { predictedLabels[idx] = true; confidences[idx] = pred.getConfidences()[i]; if (currentNode.hasChildren()) { for (HMCNode child : currentNode.getChildren()) { if (child.getName().equals(childName)) { makePrediction(child, instance, predictedLabels, confidences); } } } } else { predictedLabels[idx] = false; Set<String> descendantLabels = originalMetaData.getLabelNode(childName).getDescendantLabels(); if (descendantLabels != null) { for (String label : descendantLabels) { int idx2 = labelsAndIndices.get(label); predictedLabels[idx2] = false; confidences[idx2] = pred.getConfidences()[i]; } } } } }
From source file:mulan.classifier.meta.HOMER.java
License:Open Source License
protected MultiLabelOutput makePredictionInternal(Instance instance) throws Exception { Instance transformed = DataUtils.createInstance(instance, instance.weight(), instance.toDoubleArray()); for (int i = 0; i < numMetaLabels; i++) { transformed.insertAttributeAt(transformed.numAttributes()); }// w ww . j a va 2 s . c o m transformed.setDataset(header); MultiLabelOutput mlo = hmc.makePrediction(transformed); boolean[] oldBipartition = mlo.getBipartition(); //System.out.println("old:" + Arrays.toString(oldBipartition)); boolean[] newBipartition = new boolean[numLabels]; System.arraycopy(oldBipartition, 0, newBipartition, 0, numLabels); //System.out.println("new:" + Arrays.toString(newBipartition)); double[] oldConfidences = mlo.getConfidences(); double[] newConfidences = new double[numLabels]; System.arraycopy(oldConfidences, 0, newConfidences, 0, numLabels); MultiLabelOutput newMLO = new MultiLabelOutput(newBipartition, newConfidences); return newMLO; }
From source file:mulan.classifier.meta.thresholding.Meta.java
License:Open Source License
/** * A method that modify an instance/*from w ww . ja v a 2 s. co m*/ * * @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; }
From source file:mulan.classifier.meta.thresholding.Meta.java
License:Open Source License
/** * A method that fill the array "newValues" * * @param learner the multi-label learner * @param instance the training instances * @param newValues the array to fill//www.j a va2 s .c om * @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--; } } }
From source file:mulan.classifier.meta.thresholding.MetaLabeler.java
License:Open Source License
protected Instances transformData(MultiLabelInstances trainingData) throws Exception { // initialize classifier instances classifierInstances = RemoveAllLabels.transformInstances(trainingData); classifierInstances = new Instances(classifierInstances, 0); Attribute target = null;/*from w w w. j av a 2s . c o m*/ if (classChoice.equals("Nominal-Class")) { int countTrueLabels = 0; Set<Integer> treeSet = new TreeSet(); for (int instanceIndex = 0; instanceIndex < trainingData.getDataSet().numInstances(); instanceIndex++) { countTrueLabels = 0; for (int i = 0; i < numLabels; i++) { int labelIndice = labelIndices[i]; if (trainingData.getDataSet().attribute(labelIndice) .value((int) trainingData.getDataSet().instance(instanceIndex).value(labelIndice)) .equals("1")) { countTrueLabels++; } } treeSet.add(countTrueLabels); } ArrayList<String> classlabel = new ArrayList<String>(); for (Integer x : treeSet) { classlabel.add(x.toString()); } target = new Attribute("Class", classlabel); } else if (classChoice.equals("Numeric-Class")) { target = new Attribute("Class"); } classifierInstances.insertAttributeAt(target, classifierInstances.numAttributes()); classifierInstances.setClassIndex(classifierInstances.numAttributes() - 1); // create instances if (metaDatasetChoice.equals("Content-Based")) { for (int instanceIndex = 0; instanceIndex < trainingData.getNumInstances(); instanceIndex++) { Instance instance = trainingData.getDataSet().instance(instanceIndex); double[] values = instance.toDoubleArray(); double[] newValues = new double[classifierInstances.numAttributes()]; for (int i = 0; i < featureIndices.length; i++) { newValues[i] = values[featureIndices[i]]; } //set the number of true labels of an instance int numTrueLabels = countTrueLabels(instance); if (classChoice.compareTo("Nominal-Class") == 0) { newValues[newValues.length - 1] = classifierInstances .attribute(classifierInstances.numAttributes() - 1).indexOfValue("" + numTrueLabels); } else if (classChoice.compareTo("Numeric-Class") == 0) { newValues[newValues.length - 1] = numTrueLabels; } Instance newInstance = DataUtils.createInstance(instance, instance.weight(), newValues); classifierInstances.add(newInstance); } } else { for (int k = 0; k < kFoldsCV; k++) { //Split data to train and test sets MultiLabelLearner tempLearner; MultiLabelInstances mlTest; if (kFoldsCV == 1) { tempLearner = baseLearner; mlTest = trainingData; } else { Instances train = trainingData.getDataSet().trainCV(kFoldsCV, k); Instances test = trainingData.getDataSet().testCV(kFoldsCV, k); MultiLabelInstances mlTrain = new MultiLabelInstances(train, trainingData.getLabelsMetaData()); mlTest = new MultiLabelInstances(test, trainingData.getLabelsMetaData()); tempLearner = foldLearner.makeCopy(); tempLearner.build(mlTrain); } // copy features and labels, set metalabels for (int instanceIndex = 0; instanceIndex < mlTest.getDataSet().numInstances(); instanceIndex++) { Instance instance = mlTest.getDataSet().instance(instanceIndex); // initialize new class values double[] newValues = new double[classifierInstances.numAttributes()]; // create features valuesX(tempLearner, instance, newValues, metaDatasetChoice); //set the number of true labels of an instance int numTrueLabels = countTrueLabels(instance); if (classChoice.compareTo("Nominal-Class") == 0) { newValues[newValues.length - 1] = classifierInstances .attribute(classifierInstances.numAttributes() - 1) .indexOfValue("" + numTrueLabels); } else if (classChoice.compareTo("Numeric-Class") == 0) { newValues[newValues.length - 1] = numTrueLabels; } // add the new instance to classifierInstances Instance newInstance = DataUtils.createInstance(mlTest.getDataSet().instance(instanceIndex), mlTest.getDataSet().instance(instanceIndex).weight(), newValues); classifierInstances.add(newInstance); } } } return classifierInstances; }
From source file:mulan.classifier.neural.BPMLL.java
License:Open Source License
public MultiLabelOutput makePredictionInternal(Instance instance) throws InvalidDataException { Instance inputInstance = null;// w ww . ja va 2 s .co m if (nominalToBinaryFilter != null) { try { nominalToBinaryFilter.input(instance); inputInstance = nominalToBinaryFilter.output(); inputInstance.setDataset(null); } catch (Exception ex) { throw new InvalidDataException("The input instance for prediction is invalid. " + "Instance is not consistent with the data the model was built for."); } } else { inputInstance = DataUtils.createInstance(instance, instance.weight(), instance.toDoubleArray()); } int numAttributes = inputInstance.numAttributes(); if (numAttributes < model.getNetInputSize()) { throw new InvalidDataException("Input instance do not have enough attributes " + "to be processed by the model. Instance is not consistent with the data the model was built for."); } // if instance has more attributes than model input, we assume that true outputs // are there, so we remove them List<Integer> someLabelIndices = new ArrayList<Integer>(); boolean labelsAreThere = false; if (numAttributes > model.getNetInputSize()) { for (int index : this.labelIndices) { someLabelIndices.add(index); } labelsAreThere = true; } if (normalizeAttributes) { normalizer.normalize(inputInstance); } int inputDim = model.getNetInputSize(); double[] inputPattern = new double[inputDim]; int indexCounter = 0; for (int attrIndex = 0; attrIndex < numAttributes; attrIndex++) { if (labelsAreThere && someLabelIndices.contains(attrIndex)) { continue; } inputPattern[indexCounter] = inputInstance.value(attrIndex); indexCounter++; } double[] labelConfidences = model.feedForward(inputPattern); double threshold = thresholdF.computeThreshold(labelConfidences); boolean[] labelPredictions = new boolean[numLabels]; Arrays.fill(labelPredictions, false); for (int labelIndex = 0; labelIndex < numLabels; labelIndex++) { if (labelConfidences[labelIndex] > threshold) { labelPredictions[labelIndex] = true; } // translate from bipolar output to binary labelConfidences[labelIndex] = (labelConfidences[labelIndex] + 1) / 2; } MultiLabelOutput mlo = new MultiLabelOutput(labelPredictions, labelConfidences); return mlo; }
From source file:mulan.classifier.transformation.ClassifierChain.java
License:Open Source License
protected MultiLabelOutput makePredictionInternal(Instance instance) throws Exception { boolean[] bipartition = new boolean[numLabels]; double[] confidences = new double[numLabels]; Instance tempInstance = DataUtils.createInstance(instance, instance.weight(), instance.toDoubleArray()); for (int counter = 0; counter < numLabels; counter++) { double distribution[] = new double[2]; try {/*from www . ja v a 2 s. c o m*/ distribution = ensemble[counter].distributionForInstance(tempInstance); } catch (Exception e) { System.out.println(e); return null; } int maxIndex = (distribution[0] > distribution[1]) ? 0 : 1; // Ensure correct predictions both for class values {0,1} and {1,0} Attribute classAttribute = ensemble[counter].getFilter().getOutputFormat().classAttribute(); bipartition[chain[counter]] = (classAttribute.value(maxIndex).equals("1")) ? true : false; // The confidence of the label being equal to 1 confidences[chain[counter]] = distribution[classAttribute.indexOfValue("1")]; tempInstance.setValue(labelIndices[chain[counter]], maxIndex); } MultiLabelOutput mlo = new MultiLabelOutput(bipartition, confidences); return mlo; }
From source file:mulan.classifier.transformation.PPT.java
License:Open Source License
@Override
ArrayList<Instance> processRejected(LabelSet ls) {
switch (strategy) {
case INFORMATION_LOSS:
return new ArrayList<Instance>();
case NO_INFORMATION_LOSS:
// split LabelSet into smaller ones
//debug System.out.println("original:" + ls.toString());
ArrayList<LabelSet> subsets = null;
try {/*from w w w . j av a 2 s . co m*/
subsets = ls.getSubsets();
} catch (Exception ex) {
Logger.getLogger(PPT.class.getName()).log(Level.SEVERE, null, ex);
}
// sort subsets based on size
Collections.sort(subsets);
//debug for (LabelSet l: subsets) System.out.println(l.toString());
ArrayList<LabelSet> subsetsForInsertion = new ArrayList<LabelSet>();
for (LabelSet l : subsets) {
// check if it exists in the training set
if (!ListInstancePerLabel.containsKey(l)) {
continue;
}
// check if it occurs more than p times
if (ListInstancePerLabel.get(l).size() <= p) {
continue;
}
// check that it has no common elements with
// previously selected subsets
boolean foundCommon = false;
for (LabelSet l2 : subsetsForInsertion) {
LabelSet temp = LabelSet.intersection(l, l2);
if (temp.size() != 0) {
foundCommon = true;
break;
}
}
if (foundCommon) {
continue;
} else {
subsetsForInsertion.add(l);
}
}
// insert subsetsForInsertion with corresponding instances
// from the original labelset
ArrayList<Instance> instances = ListInstancePerLabel.get(ls);
ArrayList<Instance> newInstances = new ArrayList<Instance>();
for (Instance tempInstance : instances) {
for (LabelSet l : subsetsForInsertion) {
double[] temp = tempInstance.toDoubleArray();
double[] tempLabels = l.toDoubleArray();
for (int i = 0; i < numLabels; i++) {
if (format.attribute(labelIndices[i]).value(0).equals("0")) {
temp[labelIndices[i]] = tempLabels[i];
} else {
temp[labelIndices[i]] = 1 - tempLabels[i];
}
}
Instance newInstance = DataUtils.createInstance(tempInstance, 1, temp);
newInstances.add(newInstance);
}
}
return newInstances;
default:
return null;
}
}
From source file:mulan.classifier.transformation.PrunedSets.java
License:Open Source License
@Override
ArrayList<Instance> processRejected(LabelSet ls) {
ArrayList<LabelSet> subsets = null;
ArrayList<Instance> instances = null;
ArrayList<Instance> newInstances = null;
switch (strategy) {
case A://from ww w . j a va2 s . c o m
// split LabelSet into smaller ones
//debug System.out.println("original:" + ls.toString());
subsets = null;
try {
subsets = ls.getSubsets();
} catch (Exception ex) {
Logger.getLogger(PrunedSets.class.getName()).log(Level.SEVERE, null, ex);
}
//System.out.println("subsets: " + subsets.toString());
ArrayList<LabelSet> sortedSubsets = new ArrayList<LabelSet>();
for (LabelSet l : subsets) {
//System.out.println(l.toString());
// check if it exists in the training set
if (!ListInstancePerLabel.containsKey(l)) {
continue;
}
// check if it occurs more than p times
if (ListInstancePerLabel.get(l).size() <= p) {
continue;
}
//
boolean added = false;
for (int i = 0; i < sortedSubsets.size(); i++) {
LabelSet l2 = sortedSubsets.get(i);
if (l.size() > l2.size()) {
sortedSubsets.add(i, l);
//System.out.println("adding " + l.toString());
added = true;
break;
}
if (l.size() == l2.size()
&& ListInstancePerLabel.get(l).size() > ListInstancePerLabel.get(l2).size()) {
sortedSubsets.add(i, l);
//System.out.println("adding " + l.toString());
added = true;
break;
}
}
if (added == false) {
//System.out.println("adding " + l.toString());
sortedSubsets.add(l);
}
//System.out.println("sorted: " + sortedSubsets.toString());
}
// take the top b
newInstances = new ArrayList<Instance>();
instances = ListInstancePerLabel.get(ls);
for (Instance tempInstance : instances) {
int counter = 0;
for (LabelSet l : sortedSubsets) {
double[] temp = tempInstance.toDoubleArray();
double[] tempLabels = l.toDoubleArray();
for (int i = 0; i < numLabels; i++) {
if (format.attribute(labelIndices[i]).value(0).equals("0"))
temp[labelIndices[i]] = tempLabels[i];
else
temp[labelIndices[i]] = 1 - tempLabels[i];
}
Instance newInstance = DataUtils.createInstance(tempInstance, 1, temp);
newInstances.add(newInstance);
counter++;
if (counter == b) {
break;
}
}
}
return newInstances;
case B:
// split LabelSet into smaller ones
//debug
System.out.println("original:" + ls.toString());
subsets = null;
try {
subsets = ls.getSubsets();
} catch (Exception ex) {
Logger.getLogger(PrunedSets.class.getName()).log(Level.SEVERE, null, ex);
}
ArrayList<LabelSet> subsetsForInsertion = new ArrayList<LabelSet>();
for (LabelSet l : subsets) {
// check if it exists in the training set
if (!ListInstancePerLabel.containsKey(l)) {
continue;
}
// check if it occurs more than p times
if (ListInstancePerLabel.get(l).size() <= p) {
continue;
}
// check if it has more than b elements
if (l.size() <= b) {
continue;
}
subsetsForInsertion.add(l);
}
// insert subsetsForInsertion with corresponding instances
// from the original labelset
instances = ListInstancePerLabel.get(ls);
newInstances = new ArrayList<Instance>();
for (Instance tempInstance : instances) {
for (LabelSet l : subsetsForInsertion) {
double[] temp = tempInstance.toDoubleArray();
double[] tempLabels = l.toDoubleArray();
for (int i = 0; i < numLabels; i++) {
if (format.attribute(labelIndices[i]).value(0).equals("0"))
temp[labelIndices[i]] = tempLabels[i];
else
temp[labelIndices[i]] = 1 - tempLabels[i];
}
Instance newInstance = DataUtils.createInstance(tempInstance, 1, temp);
newInstances.add(newInstance);
}
}
System.out.println("finished");
return newInstances;
default:
return null;
}
}
From source file:mulan.regressor.transformation.RegressorChainSimple.java
License:Open Source License
protected MultiLabelOutput makePredictionInternal(Instance instance) throws Exception { double[] scores = new double[numLabels]; // create a new temporary instance so that the passed instance is not altered Instances dataset = instance.dataset(); Instance tempInstance = DataUtils.createInstance(instance, instance.weight(), instance.toDoubleArray()); for (int counter = 0; counter < numLabels; counter++) { dataset.setClassIndex(chain[counter]); tempInstance.setDataset(dataset); // find the appropriate position for that score in the scores array // i.e. which is the corresponding target int pos = 0; for (int i = 0; i < numLabels; i++) { if (chain[counter] == labelIndices[i]) { pos = i;//from www .ja v a2 s . c om break; } } scores[pos] = chainRegressors[counter].classifyInstance(tempInstance); tempInstance.setValue(chain[counter], scores[pos]); } MultiLabelOutput mlo = new MultiLabelOutput(scores, true); return mlo; }