* * Typical usage (code from the main() method of this class):
*
*
* ...
*
* // Read all the instances in the file
* reader = new FileReader(filename);
* instances = new Instances(reader);
*
* // Make the last attribute be the class
* instances.setClassIndex(instances.numAttributes() - 1);
*
* // Print header and instances.
* System.out.println("\nDataset:\n");
* System.out.println(instances);
*
* ...
*
*
* All methods that change a set of instances are safe, ie. a change
* of a set of instances does not affect any other sets of
* instances. All methods that change a datasets's attribute
* information clone the dataset before it is changed.
*
* @author Eibe Frank (eibe@cs.waikato.ac.nz)
* @author Len Trigg (trigg@cs.waikato.ac.nz)
* @version $Revision: 1.32 $
*/
public class Instances implements Serializable {
/** The filename extension that should be used for arff files */
public static String FILE_EXTENSION = ".arff";
/** The dataset's name. */
protected String m_RelationName;
/** The attribute information. */
protected FastVector m_Attributes;
/** The instances. */
protected FastVector m_Instances;
/** The class attribute's index */
protected int m_ClassIndex;
/** Buffer of values for sparse instance */
protected double[] m_ValueBuffer;
/** Buffer of indices for sparse instance */
protected int[] m_IndicesBuffer;
/**
* Reads an ARFF file from a reader, and assigns a weight of
* one to each instance. Lets the index of the class
* attribute be undefined (negative).
*
* @param reader the reader
* @exception IOException if the ARFF file is not read
* successfully
*/
public Instances(Reader reader) throws IOException {
StreamTokenizer tokenizer;
tokenizer = new StreamTokenizer(reader);
initTokenizer(tokenizer);
readHeader(tokenizer);
m_ClassIndex = -1;
m_Instances = new FastVector(1000);
while (getInstance(tokenizer, true)) {};
compactify();
}
/**
* Reads the header of an ARFF file from a reader and
* reserves space for the given number of instances. Lets
* the class index be undefined (negative).
*
* @param reader the reader
* @param capacity the capacity
* @exception IllegalArgumentException if the header is not read successfully
* or the capacity is negative.
* @exception IOException if there is a problem with the reader.
*/
public Instances(Reader reader, int capacity) throws IOException {
StreamTokenizer tokenizer;
if (capacity < 0) {
throw new IllegalArgumentException("Capacity has to be positive!");
}
tokenizer = new StreamTokenizer(reader);
initTokenizer(tokenizer);
readHeader(tokenizer);
m_ClassIndex = -1;
m_Instances = new FastVector(capacity);
}
/**
* Constructor copying all instances and references to
* the header information from the given set of instances.
*
* @param instances the set to be copied
*/
public Instances(Instances dataset) {
this(dataset, dataset.numInstances());
dataset.copyInstances(0, this, dataset.numInstances());
}
/**
* Constructor creating an empty set of instances. Copies references
* to the header information from the given set of instances. Sets
* the capacity of the set of instances to 0 if its negative.
*
* @param instances the instances from which the header
* information is to be taken
* @param capacity the capacity of the new dataset
*/
public Instances(Instances dataset, int capacity) {
if (capacity < 0) {
capacity = 0;
}
// Strings only have to be "shallow" copied because
// they can't be modified.
m_ClassIndex = dataset.m_ClassIndex;
m_RelationName = dataset.m_RelationName;
m_Attributes = dataset.m_Attributes;
m_Instances = new FastVector(capacity);
}
/**
* Creates a new set of instances by copying a
* subset of another set.
*
* @param source the set of instances from which a subset
* is to be created
* @param first the index of the first instance to be copied
* @param toCopy the number of instances to be copied
* @exception IllegalArgumentException if first and toCopy are out of range
*/
public Instances(Instances source, int first, int toCopy) {
this(source, toCopy);
if ((first < 0) || ((first + toCopy) > source.numInstances())) {
throw new IllegalArgumentException("Parameters first and/or toCopy out "+
"of range");
}
source.copyInstances(first, this, toCopy);
}
/**
* Creates an empty set of instances. Uses the given
* attribute information. Sets the capacity of the set of
* instances to 0 if its negative. Given attribute information
* must not be changed after this constructor has been used.
*
* @param name the name of the relation
* @param attInfo the attribute information
* @param capacity the capacity of the set
*/
public Instances(String name, FastVector attInfo, int capacity) {
m_RelationName = name;
m_ClassIndex = -1;
m_Attributes = attInfo;
for (int i = 0; i < numAttributes(); i++) {
attribute(i).setIndex(i);
}
m_Instances = new FastVector(capacity);
}
/**
* Create a copy of the structure, but "cleanse" string types (i.e.
* doesn't contain references to the strings seen in the past).
*
* @return a copy of the instance structure.
*/
public Instances stringFreeStructure() {
FastVector atts = (FastVector)m_Attributes.copy();
for (int i = 0 ; i < atts.size(); i++) {
Attribute att = (Attribute)atts.elementAt(i);
if (att.type() == Attribute.STRING) {
atts.setElementAt(new Attribute(att.name(), null), i);
}
}
Instances result = new Instances(relationName(), atts, 0);
result.m_ClassIndex = m_ClassIndex;
return result;
}
/**
* Adds one instance to the end of the set.
* Shallow copies instance before it is added. Increases the
* size of the dataset if it is not large enough. Does not
* check if the instance is compatible with the dataset.
*
* @param instance the instance to be added
*/
public final void add(Instance instance) {
Instance newInstance = (Instance)instance.copy();
newInstance.setDataset(this);
m_Instances.addElement(newInstance);
}
/**
* Returns an attribute.
*
* @param index the attribute's index
* @return the attribute at the given position
*/
public final Attribute attribute(int index) {
return (Attribute) m_Attributes.elementAt(index);
}
/**
* Returns an attribute given its name. If there is more than
* one attribute with the same name, it returns the first one.
* Returns null if the attribute can't be found.
*
* @param name the attribute's name
* @return the attribute with the given name, null if the
* attribute can't be found
*/
public final Attribute attribute(String name) {
for (int i = 0; i < numAttributes(); i++) {
if (attribute(i).name().equals(name)) {
return attribute(i);
}
}
return null;
}
/**
* Checks for string attributes in the dataset
*
* @return true if string attributes are present, false otherwise
*/
public boolean checkForStringAttributes() {
int i = 0;
while (i < m_Attributes.size()) {
if (attribute(i++).isString()) {
return true;
}
}
return false;
}
/**
* Checks if the given instance is compatible
* with this dataset. Only looks at the size of
* the instance and the ranges of the values for
* nominal and string attributes.
*
* @return true if the instance is compatible with the dataset
*/
public final boolean checkInstance(Instance instance) {
if (instance.numAttributes() != numAttributes()) {
return false;
}
for (int i = 0; i < numAttributes(); i++) {
if (instance.isMissing(i)) {
continue;
} else if (attribute(i).isNominal() ||
attribute(i).isString()) {
if (!(Utils.eq(instance.value(i),
(double)(int)instance.value(i)))) {
return false;
} else if (Utils.sm(instance.value(i), 0) ||
Utils.gr(instance.value(i),
attribute(i).numValues())) {
return false;
}
}
}
return true;
}
/**
* Returns the class attribute.
*
* @return the class attribute
* @exception UnassignedClassException if the class is not set
*/
public final Attribute classAttribute() {
if (m_ClassIndex < 0) {
throw new UnassignedClassException("Class index is negative (not set)!");
}
return attribute(m_ClassIndex);
}
/**
* Returns the class attribute's index. Returns negative number
* if it's undefined.
*
* @return the class index as an integer
*/
public final int classIndex() {
return m_ClassIndex;
}
/**
* Compactifies the set of instances. Decreases the capacity of
* the set so that it matches the number of instances in the set.
*/
public final void compactify() {
m_Instances.trimToSize();
}
/**
* Removes all instances from the set.
*/
public final void delete() {
m_Instances = new FastVector();
}
/**
* Removes an instance at the given position from the set.
*
* @param index the instance's position
*/
public final void delete(int index) {
m_Instances.removeElementAt(index);
}
/**
* Deletes an attribute at the given position
* (0 to numAttributes() - 1). A deep copy of the attribute
* information is performed before the attribute is deleted.
*
* @param pos the attribute's position
* @exception IllegalArgumentException if the given index is out of range or the
* class attribute is being deleted
*/
public void deleteAttributeAt(int position) {
if ((position < 0) || (position >= m_Attributes.size())) {
throw new IllegalArgumentException("Index out of range");
}
if (position == m_ClassIndex) {
throw new IllegalArgumentException("Can't delete class attribute");
}
freshAttributeInfo();
if (m_ClassIndex > position) {
m_ClassIndex--;
}
m_Attributes.removeElementAt(position);
for (int i = position; i < m_Attributes.size(); i++) {
Attribute current = (Attribute)m_Attributes.elementAt(i);
current.setIndex(current.index() - 1);
}
for (int i = 0; i < numInstances(); i++) {
instance(i).forceDeleteAttributeAt(position);
}
}
/**
* Deletes all string attributes in the dataset. A deep copy of the attribute
* information is performed before an attribute is deleted.
*
* @exception IllegalArgumentException if string attribute couldn't be
* successfully deleted (probably because it is the class attribute).
*/
public void deleteStringAttributes() {
int i = 0;
while (i < m_Attributes.size()) {
if (attribute(i).isString()) {
deleteAttributeAt(i);
} else {
i++;
}
}
}
/**
* Removes all instances with missing values for a particular
* attribute from the dataset.
*
* @param attIndex the attribute's index
*/
public final void deleteWithMissing(int attIndex) {
FastVector newInstances = new FastVector(numInstances());
for (int i = 0; i < numInstances(); i++) {
if (!instance(i).isMissing(attIndex)) {
newInstances.addElement(instance(i));
}
}
m_Instances = newInstances;
}
/**
* Removes all instances with missing values for a particular
* attribute from the dataset.
*
* @param att the attribute
*/
public final void deleteWithMissing(Attribute att) {
deleteWithMissing(att.index());
}
/**
* Removes all instances with a missing class value
* from the dataset.
*
* @exception UnassignedClassException if class is not set
*/
public final void deleteWithMissingClass() {
if (m_ClassIndex < 0) {
throw new UnassignedClassException("Class index is negative (not set)!");
}
deleteWithMissing(m_ClassIndex);
}
/**
* Returns an enumeration of all the attributes.
*
* @return enumeration of all the attributes.
*/
public Enumeration enumerateAttributes() {
return m_Attributes.elements(m_ClassIndex);
}
/**
* Returns an enumeration of all instances in the dataset.
*
* @return enumeration of all instances in the dataset
*/
public final Enumeration enumerateInstances() {
return m_Instances.elements();
}
/**
* Checks if two headers are equivalent.
*
* @param dataset another dataset
* @return true if the header of the given dataset is equivalent
* to this header
*/
public final boolean equalHeaders(Instances dataset){
// Check class and all attributes
if (m_ClassIndex != dataset.m_ClassIndex) {
return false;
}
if (m_Attributes.size() != dataset.m_Attributes.size()) {
return false;
}
for (int i = 0; i < m_Attributes.size(); i++) {
if (!(attribute(i).equals(dataset.attribute(i)))) {
return false;
}
}
return true;
}
/**
* Returns the first instance in the set.
*
* @return the first instance in the set
*/
public final Instance firstInstance() {
return (Instance)m_Instances.firstElement();
}
/**
* Inserts an attribute at the given position (0 to
* numAttributes()) and sets all values to be missing.
* Shallow copies the attribute before it is inserted, and performs
* a deep copy of the existing attribute information.
*
* @param att the attribute to be inserted
* @param pos the attribute's position
* @exception IllegalArgumentException if the given index is out of range
*/
public void insertAttributeAt(Attribute att, int position) {
if ((position < 0) ||
(position > m_Attributes.size())) {
throw new IllegalArgumentException("Index out of range");
}
att = (Attribute)att.copy();
freshAttributeInfo();
att.setIndex(position);
m_Attributes.insertElementAt(att, position);
for (int i = position + 1; i < m_Attributes.size(); i++) {
Attribute current = (Attribute)m_Attributes.elementAt(i);
current.setIndex(current.index() + 1);
}
for (int i = 0; i < numInstances(); i++) {
instance(i).forceInsertAttributeAt(position);
}
if (m_ClassIndex >= position) {
m_ClassIndex++;
}
}
/**
* Returns the instance at the given position.
*
* @param index the instance's index
* @return the instance at the given position
*/
public final Instance instance(int index) {
return (Instance)m_Instances.elementAt(index);
}
/**
* Returns the last instance in the set.
*
* @return the last instance in the set
*/
public final Instance lastInstance() {
return (Instance)m_Instances.lastElement();
}
/**
* Returns the mean (mode) for a numeric (nominal) attribute as
* a floating-point value. Returns 0 if the attribute is neither nominal nor
* numeric. If all values are missing it returns zero.
*
* @param attIndex the attribute's index
* @return the mean or the mode
*/
public final double meanOrMode(int attIndex) {
double result, found;
int [] counts;
if (attribute(attIndex).isNumeric()) {
result = found = 0;
for (int j = 0; j < numInstances(); j++) {
if (!instance(j).isMissing(attIndex)) {
found += instance(j).weight();
result += instance(j).weight()*instance(j).value(attIndex);
}
}
if (Utils.eq(found, 0)) {
return 0;
} else {
return result / found;
}
} else if (attribute(attIndex).isNominal()) {
counts = new int[attribute(attIndex).numValues()];
for (int j = 0; j < numInstances(); j++) {
if (!instance(j).isMissing(attIndex)) {
counts[(int) instance(j).value(attIndex)] += instance(j).weight();
}
}
return (double)Utils.maxIndex(counts);
} else {
return 0;
}
}
/**
* Returns the mean (mode) for a numeric (nominal) attribute as a
* floating-point value. Returns 0 if the attribute is neither
* nominal nor numeric. If all values are missing it returns zero.
*
* @param att the attribute
* @return the mean or the mode
*/
public final double meanOrMode(Attribute att) {
return meanOrMode(att.index());
}
/**
* Returns the number of attributes.
*
* @return the number of attributes as an integer
*/
public final int numAttributes() {
return m_Attributes.size();
}
/**
* Returns the number of class labels.
*
* @return the number of class labels as an integer if the class
* attribute is nominal, 1 otherwise.
* @exception UnassignedClassException if the class is not set
*/
public final int numClasses() {
if (m_ClassIndex < 0) {
throw new UnassignedClassException("Class index is negative (not set)!");
}
if (!classAttribute().isNominal()) {
return 1;
} else {
return classAttribute().numValues();
}
}
/**
* Returns the number of distinct values of a given attribute.
* Returns the number of instances if the attribute is a
* string attribute. The value 'missing' is not counted.
*
* @param attIndex the attribute
* @return the number of distinct values of a given attribute
*/
public final int numDistinctValues(int attIndex) {
if (attribute(attIndex).isNumeric()) {
double [] attVals = attributeToDoubleArray(attIndex);
int [] sorted = Utils.sort(attVals);
double prev = 0;
int counter = 0;
for (int i = 0; i < sorted.length; i++) {
Instance current = instance(sorted[i]);
if (current.isMissing(attIndex)) {
break;
}
if ((i == 0) ||
Utils.gr(current.value(attIndex), prev)) {
prev = current.value(attIndex);
counter++;
}
}
return counter;
} else {
return attribute(attIndex).numValues();
}
}
/**
* Returns the number of distinct values of a given attribute.
* Returns the number of instances if the attribute is a
* string attribute. The value 'missing' is not counted.
*
* @param att the attribute
* @return the number of distinct values of a given attribute
*/
public final int numDistinctValues(Attribute att) {
return numDistinctValues(att.index());
}
/**
* Returns the number of instances in the dataset.
*
* @return the number of instances in the dataset as an integer
*/
public final int numInstances() {
return m_Instances.size();
}
/**
* Shuffles the instances in the set so that they are ordered
* randomly.
*
* @param random a random number generator
*/
public final void randomize(Random random) {
for (int j = numInstances() - 1; j > 0; j--)
swap(j,(int)(random.nextDouble()*(double)j));
}
/**
* Reads a single instance from the reader and appends it
* to the dataset. Automatically expands the dataset if it
* is not large enough to hold the instance. This method does
* not check for carriage return at the end of the line.
*
* @param reader the reader
* @return false if end of file has been reached
* @exception IOException if the information is not read
* successfully
*/
public final boolean readInstance(Reader reader)
throws IOException {
StreamTokenizer tokenizer = new StreamTokenizer(reader);
initTokenizer(tokenizer);
return getInstance(tokenizer, false);
}
/**
* Returns the relation's name.
*
* @return the relation's name as a string
*/
public final String relationName() {
return m_RelationName;
}
/**
* Renames an attribute. This change only affects this
* dataset.
*
* @param att the attribute's index
* @param name the new name
*/
public final void renameAttribute(int att, String name) {
Attribute newAtt = attribute(att).copy(name);
FastVector newVec = new FastVector(numAttributes());
for (int i = 0; i < numAttributes(); i++) {
if (i == att) {
newVec.addElement(newAtt);
} else {
newVec.addElement(attribute(i));
}
}
m_Attributes = newVec;
}
/**
* Renames an attribute. This change only affects this
* dataset.
*
* @param att the attribute
* @param name the new name
*/
public final void renameAttribute(Attribute att, String name) {
renameAttribute(att.index(), name);
}
/**
* Renames the value of a nominal (or string) attribute value. This
* change only affects this dataset.
*
* @param att the attribute's index
* @param val the value's index
* @param name the new name
*/
public final void renameAttributeValue(int att, int val, String name) {
Attribute newAtt = (Attribute)attribute(att).copy();
FastVector newVec = new FastVector(numAttributes());
newAtt.setValue(val, name);
for (int i = 0; i < numAttributes(); i++) {
if (i == att) {
newVec.addElement(newAtt);
} else {
newVec.addElement(attribute(i));
}
}
m_Attributes = newVec;
}
/**
* Renames the value of a nominal (or string) attribute value. This
* change only affects this dataset.
*
* @param att the attribute
* @param val the value
* @param name the new name
*/
public final void renameAttributeValue(Attribute att, String val,
String name) {
renameAttributeValue(att.index(), att.indexOfValue(val), name);
}
/**
* Creates a new dataset of the same size using random sampling
* with replacement.
*
* @param random a random number generator
* @return the new dataset
*/
public final Instances resample(Random random) {
Instances newData = new Instances(this, numInstances());
while (newData.numInstances() < numInstances()) {
int i = (int) (random.nextDouble() * (double) numInstances());
newData.add(instance(i));
}
return newData;
}
/**
* Creates a new dataset of the same size using random sampling
* with replacement according to the current instance weights. The
* weights of the instances in the new dataset are set to one.
*
* @param random a random number generator
* @return the new dataset
*/
public final Instances resampleWithWeights(Random random) {
double [] weights = new double[numInstances()];
boolean foundOne = false;
for (int i = 0; i < weights.length; i++) {
weights[i] = instance(i).weight();
if (!Utils.eq(weights[i], weights[0])) {
foundOne = true;
}
}
if (foundOne) {
return resampleWithWeights(random, weights);
} else {
return new Instances(this);
}
}
/**
* Creates a new dataset of the same size using random sampling
* with replacement according to the given weight vector. The
* weights of the instances in the new dataset are set to one.
* The length of the weight vector has to be the same as the
* number of instances in the dataset, and all weights have to
* be positive.
*
* @param random a random number generator
* @param weights the weight vector
* @return the new dataset
* @exception IllegalArgumentException if the weights array is of the wrong
* length or contains negative weights.
*/
public final Instances resampleWithWeights(Random random,
double[] weights) {
if (weights.length != numInstances()) {
throw new IllegalArgumentException("weights.length != numInstances.");
}
Instances newData = new Instances(this, numInstances());
double[] probabilities = new double[numInstances()];
double sumProbs = 0, sumOfWeights = Utils.sum(weights);
for (int i = 0; i < numInstances(); i++) {
sumProbs += random.nextDouble();
probabilities[i] = sumProbs;
}
Utils.normalize(probabilities, sumProbs / sumOfWeights);
// Make sure that rounding errors don't mess things up
probabilities[numInstances() - 1] = sumOfWeights;
int k = 0; int l = 0;
sumProbs = 0;
while ((k < numInstances() && (l < numInstances()))) {
if (weights[l] < 0) {
throw new IllegalArgumentException("Weights have to be positive.");
}
sumProbs += weights[l];
while ((k < numInstances()) &&
(probabilities[k] <= sumProbs)) {
newData.add(instance(l));
newData.instance(k).setWeight(1);
k++;
}
l++;
}
return newData;
}
/**
* Sets the class attribute.
*
* @param att attribute to be the class
*/
public final void setClass(Attribute att) {
m_ClassIndex = att.index();
}
/**
* Sets the class index of the set.
* If the class index is negative there is assumed to be no class.
* (ie. it is undefined)
*
* @param classIndex the new class index
* @exception IllegalArgumentException if the class index is too big or < 0
*/
public final void setClassIndex(int classIndex) {
if (classIndex >= numAttributes()) {
throw new IllegalArgumentException("Invalid class index: " + classIndex);
}
m_ClassIndex = classIndex;
}
/**
* Sets the relation's name.
*
* @param newName the new relation name.
*/
public final void setRelationName(String newName) {
m_RelationName = newName;
}
/**
* Sorts the instances based on an attribute. For numeric attributes,
* instances are sorted in ascending order. For nominal attributes,
* instances are sorted based on the attribute label ordering
* specified in the header. Instances with missing values for the
* attribute are placed at the end of the dataset.
*
* @param attIndex the attribute's index
*/
public final void sort(int attIndex) {
int i,j;
// move all instances with missing values to end
j = numInstances() - 1;
i = 0;
while (i <= j) {
if (instance(j).isMissing(attIndex)) {
j--;
} else {
if (instance(i).isMissing(attIndex)) {
swap(i,j);
j--;
}
i++;
}
}
quickSort(attIndex, 0, j);
}
/**
* Sorts the instances based on an attribute. For numeric attributes,
* instances are sorted into ascending order. For nominal attributes,
* instances are sorted based on the attribute label ordering
* specified in the header. Instances with missing values for the
* attribute are placed at the end of the dataset.
*
* @param att the attribute
*/
public final void sort(Attribute att) {
sort(att.index());
}
/**
* Stratifies a set of instances according to its class values
* if the class attribute is nominal (so that afterwards a
* stratified cross-validation can be performed).
*
* @param numFolds the number of folds in the cross-validation
* @exception UnassignedClassException if the class is not set
*/
public final void stratify(int numFolds) {
if (numFolds <= 0) {
throw new IllegalArgumentException("Number of folds must be greater than 1");
}
if (m_ClassIndex < 0) {
throw new UnassignedClassException("Class index is negative (not set)!");
}
if (classAttribute().isNominal()) {
// sort by class
int index = 1;
while (index < numInstances()) {
Instance instance1 = instance(index - 1);
for (int j = index; j < numInstances(); j++) {
Instance instance2 = instance(j);
if ((instance1.classValue() == instance2.classValue()) ||
(instance1.classIsMissing() &&
instance2.classIsMissing())) {
swap(index,j);
index++;
}
}
index++;
}
stratStep(numFolds);
}
}
/**
* Computes the sum of all the instances' weights.
*
* @return the sum of all the instances' weights as a double
*/
public final double sumOfWeights() {
double sum = 0;
for (int i = 0; i < numInstances(); i++) {
sum += instance(i).weight();
}
return sum;
}
/**
* Creates the test set for one fold of a cross-validation on
* the dataset.
*
* @param numFolds the number of folds in the cross-validation. Must
* be greater than 1.
* @param numFold 0 for the first fold, 1 for the second, ...
* @return the test set as a set of weighted instances
* @exception IllegalArgumentException if the number of folds is less than 2
* or greater than the number of instances.
*/
public Instances testCV(int numFolds, int numFold) {
int numInstForFold, first, offset;
Instances test;
if (numFolds < 2) {
throw new IllegalArgumentException("Number of folds must be at least 2!");
}
if (numFolds > numInstances()) {
throw new IllegalArgumentException("Can't have more folds than instances!");
}
numInstForFold = numInstances() / numFolds;
if (numFold < numInstances() % numFolds){
numInstForFold++;
offset = numFold;
}else
offset = numInstances() % numFolds;
test = new Instances(this, numInstForFold);
first = numFold * (numInstances() / numFolds) + offset;
copyInstances(first, test, numInstForFold);
return test;
}
/**
* Returns the dataset as a string in ARFF format. Strings
* are quoted if they contain whitespace characters, or if they
* are a question mark.
*
* @return the dataset in ARFF format as a string
*/
public final String toString() {
StringBuffer text = new StringBuffer();
text.append("@relation " + Utils.quote(m_RelationName) + "\n\n");
for (int i = 0; i < numAttributes(); i++) {
text.append(attribute(i) + "\n");
}
text.append("\n@data\n");
for (int i = 0; i < numInstances(); i++) {
text.append(instance(i));
if (i < numInstances() - 1) {
text.append('\n');
}
}
return text.toString();
}
/**
* Creates the training set for one fold of a cross-validation
* on the dataset.
*
* @param numFolds the number of folds in the cross-validation. Must
* be greater than 1.
* @param numFold 0 for the first fold, 1 for the second, ...
* @return the training set as a set of weighted
* instances
* @exception IllegalArgumentException if the number of folds is less than 2
* or greater than the number of instances.
*/
public Instances trainCV(int numFolds, int numFold) {
int numInstForFold, first, offset;
Instances train;
if (numFolds < 2) {
throw new IllegalArgumentException("Number of folds must be at least 2!");
}
if (numFolds > numInstances()) {
throw new IllegalArgumentException("Can't have more folds than instances!");
}
numInstForFold = numInstances() / numFolds;
if (numFold < numInstances() % numFolds) {
numInstForFold++;
offset = numFold;
}else
offset = numInstances() % numFolds;
train = new Instances(this, numInstances() - numInstForFold);
first = numFold * (numInstances() / numFolds) + offset;
copyInstances(0, train, first);
copyInstances(first + numInstForFold, train,
numInstances() - first - numInstForFold);
return train;
}
/**
* Computes the variance for a numeric attribute.
*
* @param attIndex the numeric attribute
* @return the variance if the attribute is numeric
* @exception IllegalArgumentException if the attribute is not numeric
*/
public final double variance(int attIndex) {
double sum = 0, sumSquared = 0, sumOfWeights = 0;
if (!attribute(attIndex).isNumeric()) {
throw new IllegalArgumentException("Can't compute variance because attribute is " +
"not numeric!");
}
for (int i = 0; i < numInstances(); i++) {
if (!instance(i).isMissing(attIndex)) {
sum += instance(i).weight() *
instance(i).value(attIndex);
sumSquared += instance(i).weight() *
instance(i).value(attIndex) *
instance(i).value(attIndex);
sumOfWeights += instance(i).weight();
}
}
if (Utils.smOrEq(sumOfWeights, 1)) {
return 0;
}
return (sumSquared - (sum * sum / sumOfWeights)) /
(sumOfWeights - 1);
}
/**
* Computes the variance for a numeric attribute.
*
* @param att the numeric attribute
* @return the variance if the attribute is numeric
* @exception IllegalArgumentException if the attribute is not numeric
*/
public final double variance(Attribute att) {
return variance(att.index());
}
/**
* Calculates summary statistics on the values that appear in this
* set of instances for a specified attribute.
*
* @param index the index of the attribute to summarize.
* @return an AttributeStats object with it's fields calculated.
*/
public AttributeStats attributeStats(int index) {
AttributeStats result = new AttributeStats();
if (attribute(index).isNominal()) {
result.nominalCounts = new int [attribute(index).numValues()];
}
if (attribute(index).isNumeric()) {
result.numericStats = new weka.experiment.Stats();
}
result.totalCount = numInstances();
double [] attVals = attributeToDoubleArray(index);
int [] sorted = Utils.sort(attVals);
int currentCount = 0;
double prev = Instance.missingValue();
for (int j = 0; j < numInstances(); j++) {
Instance current = instance(sorted[j]);
if (current.isMissing(index)) {
result.missingCount = numInstances() - j;
break;
}
if (Utils.eq(current.value(index), prev)) {
currentCount++;
} else {
result.addDistinct(prev, currentCount);
currentCount = 1;
prev = current.value(index);
}
}
result.addDistinct(prev, currentCount);
result.distinctCount--; // So we don't count "missing" as a value
return result;
}
/**
* Gets the value of all instances in this dataset for a particular
* attribute. Useful in conjunction with Utils.sort to allow iterating
* through the dataset in sorted order for some attribute.
*
* @param index the index of the attribute.
* @return an array containing the value of the desired attribute for
* each instance in the dataset.
*/
public double [] attributeToDoubleArray(int index) {
double [] result = new double[numInstances()];
for (int i = 0; i < result.length; i++) {
result[i] = instance(i).value(index);
}
return result;
}
/**
* Generates a string summarizing the set of instances. Gives a breakdown
* for each attribute indicating the number of missing/discrete/unique
* values and other information.
*
* @return a string summarizing the dataset
*/
public String toSummaryString() {
StringBuffer result = new StringBuffer();
result.append("Relation Name: ").append(relationName()).append('\n');
result.append("Num Instances: ").append(numInstances()).append('\n');
result.append("Num Attributes: ").append(numAttributes()).append('\n');
result.append('\n');
result.append(Utils.padLeft("", 5)).append(Utils.padRight("Name", 25));
result.append(Utils.padLeft("Type", 5)).append(Utils.padLeft("Nom", 5));
result.append(Utils.padLeft("Int", 5)).append(Utils.padLeft("Real", 5));
result.append(Utils.padLeft("Missing", 12));
result.append(Utils.padLeft("Unique", 12));
result.append(Utils.padLeft("Dist", 6)).append('\n');
for (int i = 0; i < numAttributes(); i++) {
Attribute a = attribute(i);
AttributeStats as = attributeStats(i);
result.append(Utils.padLeft("" + (i + 1), 4)).append(' ');
result.append(Utils.padRight(a.name(), 25)).append(' ');
long percent;
switch (a.type()) {
case Attribute.NOMINAL:
result.append(Utils.padLeft("Nom", 4)).append(' ');
percent = Math.round(100.0 * as.intCount / as.totalCount);
result.append(Utils.padLeft("" + percent, 3)).append("% ");
result.append(Utils.padLeft("" + 0, 3)).append("% ");
percent = Math.round(100.0 * as.realCount / as.totalCount);
result.append(Utils.padLeft("" + percent, 3)).append("% ");
break;
case Attribute.NUMERIC:
result.append(Utils.padLeft("Num", 4)).append(' ');
result.append(Utils.padLeft("" + 0, 3)).append("% ");
percent = Math.round(100.0 * as.intCount / as.totalCount);
result.append(Utils.padLeft("" + percent, 3)).append("% ");
percent = Math.round(100.0 * as.realCount / as.totalCount);
result.append(Utils.padLeft("" + percent, 3)).append("% ");
break;
case Attribute.STRING:
result.append(Utils.padLeft("Str", 4)).append(' ');
percent = Math.round(100.0 * as.intCount / as.totalCount);
result.append(Utils.padLeft("" + percent, 3)).append("% ");
result.append(Utils.padLeft("" + 0, 3)).append("% ");
percent = Math.round(100.0 * as.realCount / as.totalCount);
result.append(Utils.padLeft("" + percent, 3)).append("% ");
break;
default:
result.append(Utils.padLeft("???", 4)).append(' ');
result.append(Utils.padLeft("" + 0, 3)).append("% ");
percent = Math.round(100.0 * as.intCount / as.totalCount);
result.append(Utils.padLeft("" + percent, 3)).append("% ");
percent = Math.round(100.0 * as.realCount / as.totalCount);
result.append(Utils.padLeft("" + percent, 3)).append("% ");
break;
}
result.append(Utils.padLeft("" + as.missingCount, 5)).append(" /");
percent = Math.round(100.0 * as.missingCount / as.totalCount);
result.append(Utils.padLeft("" + percent, 3)).append("% ");
result.append(Utils.padLeft("" + as.uniqueCount, 5)).append(" /");
percent = Math.round(100.0 * as.uniqueCount / as.totalCount);
result.append(Utils.padLeft("" + percent, 3)).append("% ");
result.append(Utils.padLeft("" + as.distinctCount, 5)).append(' ');
result.append('\n');
}
return result.toString();
}
/**
* Reads a single instance using the tokenizer and appends it
* to the dataset. Automatically expands the dataset if it
* is not large enough to hold the instance.
*
* @param tokenizer the tokenizer to be used
* @param flag if method should test for carriage return after
* each instance
* @return false if end of file has been reached
* @exception IOException if the information is not read
* successfully
*/
protected boolean getInstance(StreamTokenizer tokenizer,
boolean flag)
throws IOException {
// Check if any attributes have been declared.
if (m_Attributes.size() == 0) {
errms(tokenizer,"no header information available");
}
// Check if end of file reached.
getFirstToken(tokenizer);
if (tokenizer.ttype == StreamTokenizer.TT_EOF) {
return false;
}
// Parse instance
if (tokenizer.ttype == '{') {
return getInstanceSparse(tokenizer, flag);
} else {
return getInstanceFull(tokenizer, flag);
}
}
/**
* Reads a single instance using the tokenizer and appends it
* to the dataset. Automatically expands the dataset if it
* is not large enough to hold the instance.
*
* @param tokenizer the tokenizer to be used
* @param flag if method should test for carriage return after
* each instance
* @return false if end of file has been reached
* @exception IOException if the information is not read
* successfully
*/
protected boolean getInstanceSparse(StreamTokenizer tokenizer,
boolean flag)
throws IOException {
int valIndex, numValues = 0, maxIndex = -1;
// Get values
do {
// Get index
getIndex(tokenizer);
if (tokenizer.ttype == '}') {
break;
}
// Is index valid?
try{
m_IndicesBuffer[numValues] = Integer.valueOf(tokenizer.sval).intValue();
} catch (NumberFormatException e) {
errms(tokenizer,"index number expected");
}
if (m_IndicesBuffer[numValues] <= maxIndex) {
errms(tokenizer,"indices have to be ordered");
}
if ((m_IndicesBuffer[numValues] < 0) ||
(m_IndicesBuffer[numValues] >= numAttributes())) {
errms(tokenizer,"index out of bounds");
}
maxIndex = m_IndicesBuffer[numValues];
// Get value;
getNextToken(tokenizer);
// Check if value is missing.
if (tokenizer.ttype == '?') {
m_ValueBuffer[numValues] = Instance.missingValue();
} else {
// Check if token is valid.
if (tokenizer.ttype != StreamTokenizer.TT_WORD) {
errms(tokenizer,"not a valid value");
}
if (attribute(m_IndicesBuffer[numValues]).isNominal()) {
// Check if value appears in header.
valIndex =
attribute(m_IndicesBuffer[numValues]).indexOfValue(tokenizer.sval);
if (valIndex == -1) {
errms(tokenizer,"nominal value not declared in header");
}
m_ValueBuffer[numValues] = (double)valIndex;
} else if (attribute(m_IndicesBuffer[numValues]).isNumeric()) {
// Check if value is really a number.
try{
m_ValueBuffer[numValues] = Double.valueOf(tokenizer.sval).
doubleValue();
} catch (NumberFormatException e) {
errms(tokenizer,"number expected");
}
} else {
m_ValueBuffer[numValues] =
attribute(m_IndicesBuffer[numValues]).addStringValue(tokenizer.sval);
}
}
numValues++;
} while (true);
if (flag) {
getLastToken(tokenizer,true);
}
// Add instance to dataset
double[] tempValues = new double[numValues];
int[] tempIndices = new int[numValues];
System.arraycopy(m_ValueBuffer, 0, tempValues, 0, numValues);
System.arraycopy(m_IndicesBuffer, 0, tempIndices, 0, numValues);
add(new SparseInstance(1, tempValues, tempIndices, numAttributes()));
return true;
}
/**
* Reads a single instance using the tokenizer and appends it
* to the dataset. Automatically expands the dataset if it
* is not large enough to hold the instance.
*
* @param tokenizer the tokenizer to be used
* @param flag if method should test for carriage return after
* each instance
* @return false if end of file has been reached
* @exception IOException if the information is not read
* successfully
*/
protected boolean getInstanceFull(StreamTokenizer tokenizer,
boolean flag)
throws IOException {
double[] instance = new double[numAttributes()];
int index;
// Get values for all attributes.
for (int i = 0; i < numAttributes(); i++){
// Get next token
if (i > 0) {
getNextToken(tokenizer);
}
// Check if value is missing.
if (tokenizer.ttype == '?') {
instance[i] = Instance.missingValue();
} else {
// Check if token is valid.
if (tokenizer.ttype != StreamTokenizer.TT_WORD) {
errms(tokenizer,"not a valid value");
}
if (attribute(i).isNominal()) {
// Check if value appears in header.
index = attribute(i).indexOfValue(tokenizer.sval);
if (index == -1) {
errms(tokenizer,"nominal value not declared in header");
}
instance[i] = (double)index;
} else if (attribute(i).isNumeric()) {
// Check if value is really a number.
try{
instance[i] = Double.valueOf(tokenizer.sval).
doubleValue();
} catch (NumberFormatException e) {
errms(tokenizer,"number expected");
}
} else {
instance[i] = attribute(i).addStringValue(tokenizer.sval);
}
}
}
if (flag) {
getLastToken(tokenizer,true);
}
// Add instance to dataset
add(new Instance(1, instance));
return true;
}
/**
* Reads and stores header of an ARFF file.
*
* @param tokenizer the stream tokenizer
* @exception IOException if the information is not read
* successfully
*/
protected void readHeader(StreamTokenizer tokenizer)
throws IOException{
String attributeName;
FastVector attributeValues;
int i;
// Get name of relation.
getFirstToken(tokenizer);
if (tokenizer.ttype == StreamTokenizer.TT_EOF) {
errms(tokenizer,"premature end of file");
}
if (tokenizer.sval.equalsIgnoreCase("@relation")){
getNextToken(tokenizer);
m_RelationName = tokenizer.sval;
getLastToken(tokenizer,false);
} else {
errms(tokenizer,"keyword @relation expected");
}
// Create vectors to hold information temporarily.
m_Attributes = new FastVector();
// Get attribute declarations.
getFirstToken(tokenizer);
if (tokenizer.ttype == StreamTokenizer.TT_EOF) {
errms(tokenizer,"premature end of file");
}
while (tokenizer.sval.equalsIgnoreCase("@attribute")) {
// Get attribute name.
getNextToken(tokenizer);
attributeName = tokenizer.sval;
getNextToken(tokenizer);
// Check if attribute is nominal.
if (tokenizer.ttype == StreamTokenizer.TT_WORD) {
// Attribute is real, integer, or string.
if (tokenizer.sval.equalsIgnoreCase("real") ||
tokenizer.sval.equalsIgnoreCase("integer") ||
tokenizer.sval.equalsIgnoreCase("numeric")) {
m_Attributes.addElement(new Attribute(attributeName,
numAttributes()));
readTillEOL(tokenizer);
} else if (tokenizer.sval.equalsIgnoreCase("string")) {
m_Attributes.
addElement(new Attribute(attributeName, null,
numAttributes()));
readTillEOL(tokenizer);
} else {
errms(tokenizer,"no valid attribute type or invalid "+
"enumeration");
}
} else {
// Attribute is nominal.
attributeValues = new FastVector();
tokenizer.pushBack();
// Get values for nominal attribute.
if (tokenizer.nextToken() != '{') {
errms(tokenizer,"{ expected at beginning of enumeration");
}
while (tokenizer.nextToken() != '}') {
if (tokenizer.ttype == StreamTokenizer.TT_EOL) {
errms(tokenizer,"} expected at end of enumeration");
} else {
attributeValues.addElement(tokenizer.sval);
}
}
if (attributeValues.size() == 0) {
errms(tokenizer,"no nominal values found");
}
m_Attributes.
addElement(new Attribute(attributeName, attributeValues,
numAttributes()));
}
getLastToken(tokenizer,false);
getFirstToken(tokenizer);
if (tokenizer.ttype == StreamTokenizer.TT_EOF)
errms(tokenizer,"premature end of file");
}
// Check if data part follows. We can't easily check for EOL.
if (!tokenizer.sval.equalsIgnoreCase("@data")) {
errms(tokenizer,"keyword @data expected");
}
// Check if any attributes have been declared.
if (m_Attributes.size() == 0) {
errms(tokenizer,"no attributes declared");
}
// Allocate buffers in case sparse instances have to be read
m_ValueBuffer = new double[numAttributes()];
m_IndicesBuffer = new int[numAttributes()];
}
/**
* Copies instances from one set to the end of another
* one.
*
* @param source the source of the instances
* @param from the position of the first instance to be copied
* @param dest the destination for the instances
* @param num the number of instances to be copied
*/
private void copyInstances(int from, Instances dest, int num) {
for (int i = 0; i < num; i++) {
dest.add(instance(from + i));
}
}
/**
* Throws error message with line number and last token read.
*
* @param theMsg the error message to be thrown
* @param tokenizer the stream tokenizer
* @throws IOExcpetion containing the error message
*/
private void errms(StreamTokenizer tokenizer, String theMsg)
throws IOException {
throw new IOException(theMsg + ", read " + tokenizer.toString());
}
/**
* Replaces the attribute information by a clone of
* itself.
*/
private void freshAttributeInfo() {
m_Attributes = (FastVector) m_Attributes.copyElements();
}
/**
* Gets next token, skipping empty lines.
*
* @param tokenizer the stream tokenizer
* @exception IOException if reading the next token fails
*/
private void getFirstToken(StreamTokenizer tokenizer)
throws IOException{
while (tokenizer.nextToken() == StreamTokenizer.TT_EOL){};
if ((tokenizer.ttype == '\'') ||
(tokenizer.ttype == '"')) {
tokenizer.ttype = StreamTokenizer.TT_WORD;
} else if ((tokenizer.ttype == StreamTokenizer.TT_WORD) &&
(tokenizer.sval.equals("?"))){
tokenizer.ttype = '?';
}
}
/**
* Gets index, checking for a premature and of line.
*
* @param tokenizer the stream tokenizer
* @exception IOException if it finds a premature end of line
*/
private void getIndex(StreamTokenizer tokenizer) throws IOException{
if (tokenizer.nextToken() == StreamTokenizer.TT_EOL) {
errms(tokenizer,"premature end of line");
}
if (tokenizer.ttype == StreamTokenizer.TT_EOF) {
errms(tokenizer,"premature end of file");
}
}
/**
* Gets token and checks if its end of line.
*
* @param tokenizer the stream tokenizer
* @exception IOException if it doesn't find an end of line
*/
private void getLastToken(StreamTokenizer tokenizer, boolean endOfFileOk)
throws IOException{
if ((tokenizer.nextToken() != StreamTokenizer.TT_EOL) &&
((tokenizer.nextToken() != StreamTokenizer.TT_EOF) || !endOfFileOk)) {
errms(tokenizer,"end of line expected");
}
}
/**
* Gets next token, checking for a premature and of line.
*
* @param tokenizer the stream tokenizer
* @exception IOException if it finds a premature end of line
*/
private void getNextToken(StreamTokenizer tokenizer)
throws IOException{
if (tokenizer.nextToken() == StreamTokenizer.TT_EOL) {
errms(tokenizer,"premature end of line");
}
if (tokenizer.ttype == StreamTokenizer.TT_EOF) {
errms(tokenizer,"premature end of file");
} else if ((tokenizer.ttype == '\'') ||
(tokenizer.ttype == '"')) {
tokenizer.ttype = StreamTokenizer.TT_WORD;
} else if ((tokenizer.ttype == StreamTokenizer.TT_WORD) &&
(tokenizer.sval.equals("?"))){
tokenizer.ttype = '?';
}
}
/**
* Initializes the StreamTokenizer used for reading the ARFF file.
*
* @param tokenizer the stream tokenizer
*/
private void initTokenizer(StreamTokenizer tokenizer){
tokenizer.resetSyntax();
tokenizer.whitespaceChars(0, ' ');
tokenizer.wordChars(' '+1,'\u00FF');
tokenizer.whitespaceChars(',',',');
tokenizer.commentChar('%');
tokenizer.quoteChar('"');
tokenizer.quoteChar('\'');
tokenizer.ordinaryChar('{');
tokenizer.ordinaryChar('}');
tokenizer.eolIsSignificant(true);
}
/**
* Returns string including all instances, their weights and
* their indices in the original dataset.
*
* @return description of instance and its weight as a string
*/
private String instancesAndWeights(){
StringBuffer text = new StringBuffer();
for (int i = 0; i < numInstances(); i++) {
text.append(instance(i) + " " + instance(i).weight());
if (i < numInstances() - 1) {
text.append("\n");
}
}
return text.toString();
}
/**
* Implements quicksort.
*
* @param attIndex the attribute's index
* @param lo0 the first index of the subset to be sorted
* @param hi0 the last index of the subset to be sorted
*/
private void quickSort(int attIndex, int lo0, int hi0) {
int lo = lo0, hi = hi0;
double mid, midPlus, midMinus;
if (hi0 > lo0) {
// Arbitrarily establishing partition element as the
// midpoint of the array.
mid = instance((lo0 + hi0) / 2).value(attIndex);
midPlus = mid + 1e-6;
midMinus = mid - 1e-6;
// loop through the array until indices cross
while(lo <= hi) {
// find the first element that is greater than or equal to
// the partition element starting from the left Index.
while ((instance(lo).value(attIndex) <
midMinus) && (lo < hi0)) {
++lo;
}
// find an element that is smaller than or equal to
// the partition element starting from the right Index.
while ((instance(hi).value(attIndex) >
midPlus) && (hi > lo0)) {
--hi;
}
// if the indexes have not crossed, swap
if(lo <= hi) {
swap(lo,hi);
++lo;
--hi;
}
}
// If the right index has not reached the left side of array
// must now sort the left partition.
if(lo0 < hi) {
quickSort(attIndex,lo0,hi);
}
// If the left index has not reached the right side of array
// must now sort the right partition.
if(lo < hi0) {
quickSort(attIndex,lo,hi0);
}
}
}
/**
* Reads and skips all tokens before next end of line token.
*
* @param tokenizer the stream tokenizer
*/
private void readTillEOL(StreamTokenizer tokenizer)
throws IOException{
while (tokenizer.nextToken() != StreamTokenizer.TT_EOL) {};
tokenizer.pushBack();
}
/**
* Help function needed for stratification of set.
*
* @param numFolds the number of folds for the stratification
*/
private void stratStep (int numFolds){
FastVector newVec = new FastVector(m_Instances.capacity());
int start = 0, j;
// create stratified batch
while (newVec.size() < numInstances()) {
j = start;
while (j < numInstances()) {
newVec.addElement(instance(j));
j = j + numFolds;
}
start++;
}
m_Instances = newVec;
}
/**
* Swaps two instances in the set.
*
* @param i the first instance's index
* @param j the second instance's index
*/
private void swap(int i, int j){
m_Instances.swap(i, j);
}
/**
* Merges two sets of Instances together. The resulting set will have
* all the attributes of the first set plus all the attributes of the
* second set. The number of instances in both sets must be the same.
*
* @param first the first set of Instances
* @param second the second set of Instances
* @return the merged set of Instances
* @exception IllegalArgumentException if the datasets are not the same size
*/
public static Instances mergeInstances(Instances first, Instances second) {
if (first.numInstances() != second.numInstances()) {
throw new IllegalArgumentException("Instance sets must be of the same size");
}
// Create the vector of merged attributes
FastVector newAttributes = new FastVector();
for (int i = 0; i < first.numAttributes(); i++) {
newAttributes.addElement(first.attribute(i));
}
for (int i = 0; i < second.numAttributes(); i++) {
newAttributes.addElement(second.attribute(i));
}
// Create the set of Instances
Instances merged = new Instances(first.relationName() + '_'
+ second.relationName(),
newAttributes,
first.numInstances());
// Merge each instance
for (int i = 0; i < first.numInstances(); i++) {
merged.add(first.instance(i).mergeInstance(second.instance(i)));
}
return merged;
}
/**
* Method for testing this class.
*
* @param argv should contain one element: the name of an ARFF file
*/
public static void test(String [] argv) {
Instances instances, secondInstances, train, test, transformed, empty;
Instance instance;
Random random = new Random(2);
Reader reader;
int start, num;
double newWeight;
FastVector testAtts, testVals;
int i,j;
try{
if (argv.length > 1) {
throw (new Exception("Usage: Instances [