API Documentation¶
This section contains the documentation of all classes and functions of MOAFS.
Naïve Bayes¶
class moa.featureselection.classifiers.NaiveBayes()
This is an incremental version of the Naïve Bayes classifier. This is the main class of the MOAFS library. It receives the requests from MOA and performs
prediction using (or not) the feature selection methods available at MOAFS. Since it is using the Interleaved-Test-Then-Train approach, each instance is used
to test the model before it is used for training.
public String getPurposeString()
Method used by MOA to display the class purpose at the GUI.
- Parameters
- None
- Returns
- A string explaining the class purpose.
public boolean isRandomizable()
Necessary method from the AbstractClassifier class. It Gets whether this learner needs a random seed. Examples of methods that needs a random seed are bagging and boosting.
- Parameters
- None
- Returns
- False
public void getModelDescription(StringBuilder out, int indent)
Necessary method from the AbstractClassifier class. Returns a string representation of the model.
- Parameters
- StringBuilder out: A mutable sequence of characters. This class provides an API compatible with StringBuffer, but with no guarantee of synchronization.
- int indent: The number of characters to indent.
- Returns
- None
protected Measurement[] getModelMeasurementsImpl()
Necessary method from the AbstractClassifier class. Class for storing an evaluation measurement.
- Parameters
- None
- Returns
- Null
public double[] getVotesForInstance(com.yahoo.labs.samoa.instances.Instance inst)
Necessary method from the AbstractClassifier class. Predicts the class memberships for a given instance. If an instance is unclassified, the returned array elements must be all zero.
- Parameters
- inst : The instance to be classified.
- Returns
- The prediction of the instance class using doNaiveBayesPrediction() method.
public void resetLearningImpl()
Necessary method from the AbstractClassifier class. Resets this classifier. It must be similar to starting a new classifier from scratch.
- Parameters
- None
- Returns
- None
public void trainOnInstanceImpl(com.yahoo.labs.samoa.instances.Instance inst)
Necessary method from the AbstractClassifier class. Trains this classifier incrementally using the given instance.
- Parameters
- inst : The instance to be classified.
- Returns
- None
protected AttributeClassObserver newNominalClassObserver()
Interface for observing the class data distribution for an attribute. This observer monitors the class distribution of a given attribute. Used in naive Bayes and decision trees to monitor data statistics on leaves.
- Parameters
- None
- Returns
- NominalAttributeClassObserver(): Class distribution of a given nominal attribute.
protected AttributeClassObserver newNumericClassObserver()
Interface for observing the class data distribution for an attribute. This observer monitors the class distribution of a given attribute. Used in naive Bayes and decision trees to monitor data statistics on leaves.
- Parameters
- None
- Returns
- new GaussianNumericAttributeClassObserver(): Class distribution of a given numeric attribute.
private void performFeatureSelection(com.yahoo.labs.samoa.instances.Instance rinst)
Performs Feature Selection using the selected methods.
- Parameters
- rinst : The instance to be reduced.
- Returns
- None
Information-based algorithms¶
class moa.featureselection.algorithms.IncrInfoThAttributeEval()
Evaluates the worth of an attribute by measuring the information gain, chi-squared, symmetrical uncertainty, crammer’s v-test or gain ratio with respect to the class.
public String globalInfo()
Returns a string describing this attribute evaluator.
- Parameters
- numFeatures: Number of features to select.
- Returns
- A description of the evaluator suitable for displaying in the explorer/experimenter gui
public String binarizeNumericAttributesTipText()
Returns the tip text for this property
- Parameters
- numFeatures: Number of features to select.
- Returns
- A tip text for this property suitable for displaying in the explorer/experimenter gui
public void setBinarizeNumericAttributes(boolean b)
Binarize numeric attributes.
- Parameters
- b: if true, binarize numeric attributes
- Returns
- None
public boolean getBinarizeNumericAttributes()
Get whether numeric attributes are just being binarized.
- Parameters
- None
- Returns
- True if missing values are being distributed.
public String missingMergeTipText()
Returns the tip text for this property.
- Parameters
- None
- Returns
- Tip text for this property suitable for displaying in the explorer/experimenter gui
public void setMissingMerge(boolean b)
Distribute the counts for missing values across observed values
- Parameters
- b: if true, distribute missing values.
- Returns
- None
public boolean getMissingMerge()
Get whether missing values are being distributed or not
- Parameters
- None
- Returns
- True if missing values are being distributed.
public Capabilities getCapabilities()
Returns the capabilities of this evaluator.
- Parameters
- None
- Returns
- The capabilities of this evaluator
public void buildEvaluator(weka.core.Instances data)
Initializes an information gain attribute evaluator. Discretizes all attributes that are numeric.
- Parameters
- data: data set of instances serving as training data
- Returns
- None
public void updateEvaluator(Instance inst)
Updates an information gain attribute evaluator. Discretizes all attributes that are numeric.
- Parameters
- dnst: data set of instances serving as training data
- Returns
- None
public void applySelection()
Update the contingency tables and the rankings for each features using the counters. Counters are updated in each iteration.
- Parameters
- None
- Returns
- None
public String toString()
Describe the attribute evaluator
- Parameters
- None
- Returns
- A description of the attribute evaluator as a string
public String getRevision()
Returns the revision string.
- Parameters
- None
- Returns
- The revision
Attribute Evaluator¶
Interface class to guarantee all classes contains the necessary methods.
public void updateEvaluator(Instance inst)
Updates an attribute evaluator using the specified approach.
public void applySelection()
Used to verify if a selection is needed.
public boolean isUpdated()
Verifies if the attribute evaluator is upated with the new instance.
Extremal Feature Selection¶
class moa.featureselection.algorithms.ExtremalFeatureSelection()
Evaluates the worth of an attribute through the computation of weights based on the Modified Balanced Winnow. Original paper on https://dl.acm.org/citation.cfm?doid=1150402.1150466.
public ExtremalFeatureSelection(int numFeatures)
Constructor.
- Parameters
- numFeatures: Number of features to select.
- Returns
- None
public String globalInfo()
Returns a string describing this attribute evaluator
- Parameters
- None
- Returns
- A description of the evaluator suitable for displaying in the explorer/experimenter gui
public String toString()
Describe the attribute evaluator
- Parameters
- None
- Returns
- A description of the attribute evaluator as a string
public String toString()
Describe the attribute evaluator
- Parameters
- None
- Returns
- A description of the attribute evaluator as a string
public static void getMaxMinValue(double[] numbers)
Get both maximum and minimum values of a given instance
- Parameters
- numbers: Instance values
- Returns
- None
public double[][] rankedAttributes(int[] m_attributeList, double[] m_attributeMerit)
Sorts instance’s attribute values from best to worst
- Parameters
- m_attributeList: Index of every attribute value
- m_attributeMerit: Attribute values
- Returns
- None
findLowestScore()
Find the lowest scores to include in the selected features according to authors criteria
- Parameters
- None
- Returns
- None
public void updateModels(double[] normalizedData, double trueClass)
Necessary method from the ASEvalution class. Updates the positive and negative models.
- Parameters
- normalizedData: Instance normalized data
- trueClass: Instance true class
- Returns
- None
public void updateEvaluator(Instance inst)
Necessary method from the MOAAttributeEvalutor interface. Updates the evaluator with the new information.
- Parameters
- inst: The instance to be classified.
- Returns
- None
public void applySelection()
Used to verify if a selection is needed.
- Parameters
- None
- Returns
- None
public double evaluateAttribute(int attribute)
Evaluate attribute relevance.
- Parameters
- attribute: The index of the attribute to be evaluated.
- Returns
- None
Online Feature Selection¶
class moa.featureselection.algorithms.OFS()
Evaluates the worth of an attribute through the computation of weights using a linear classifier with sparse projection. Original paper on https://doi.ieeecomputersociety.org/10.1109/TKDE.2013.32.
public String globalInfo()
Returns a string describing this attribute evaluator.
- Parameters
- numFeatures: Number of features to select.
- Returns
- A description of the evaluator suitable for displaying in the explorer/experimenter gui
public String binarizeNumericAttributesTipText()
Returns the tip text for this property
- Parameters
- numFeatures: Number of features to select.
- Returns
- A tip text for this property suitable for displaying in the explorer/experimenter gui
public void setBinarizeNumericAttributes(boolean b)
Binarize numeric attributes.
- Parameters
- b: if true, binarize numeric attributes
- Returns
- None
public boolean getBinarizeNumericAttributes()
Get whether numeric attributes are just being binarized.
- Parameters
- None
- Returns
- True if missing values are being distributed.
public String missingMergeTipText()
Returns the tip text for this property.
- Parameters
- None
- Returns
- Tip text for this property suitable for displaying in the explorer/experimenter gui
public void setMissingMerge(boolean b)
Distribute the counts for missing values across observed values
- Parameters
- b: if true, distribute missing values.
- Returns
- None
public boolean getMissingMerge()
Get whether missing values are being distributed or not
- Parameters
- None
- Returns
- True if missing values are being distributed.
public Capabilities getCapabilities()
Returns the capabilities of this evaluator.
- Parameters
- None
- Returns
- The capabilities of this evaluator
public void buildEvaluator(weka.core.Instances data)
Initializes an sparse projection attribute evaluator.
- Parameters
- data: data set of instances serving as training data
- Returns
- None
public void updateEvaluator(Instance inst)
Updates an attribute evaluator using the sparse projection approach.
- Parameters
- dnst: data set of instances serving as training data
- Returns
- None
public void applySelection()
Used to verify if a selection is needed.
- Parameters
- None
- Returns
- None
public String toString()
Describe the attribute evaluator
- Parameters
- None
- Returns
- A description of the attribute evaluator as a string
public String getRevision()
Returns the revision string.
- Parameters
- None
- Returns
- The revision