/**
* Generate artificial training examples.
*
* @param artSize size of examples set to create
* @param data training data
* @return the set of unlabeled artificial examples
*/
protected Instances generateArtificialData(int artSize, Instances data) {
int numAttributes = data.numAttributes();
Instances artData = new Instances(data, artSize);
double[] att;
Instance artInstance;
for (int i = 0; i < artSize; i++) {
att = new double[numAttributes];
for (int j = 0; j < numAttributes; j++) {
if (data.attribute(j).isNominal()) {
// Select nominal value based on the frequency of occurence in the training data
double[] stats = (double[]) m_AttributeStats.get(j);
att[j] = (double) selectIndexProbabilistically(stats);
} else if (data.attribute(j).isNumeric()) {
// Generate numeric value from the Guassian distribution
// defined by the mean and std dev of the attribute
double[] stats = (double[]) m_AttributeStats.get(j);
att[j] = (m_Random.nextGaussian() * stats[1]) + stats[0];
} else System.err.println("Decorate can only handle numeric and nominal values.");
}
artInstance = new Instance(1.0, att);
artData.add(artInstance);
}
return artData;
}
// 构造一个tri-trainer分类器。
public Tritrainer(
String classifier, String trainingIns_File, String testIns_File, double precentage) {
try {
this.classifier1 = (Classifier) Class.forName(classifier).newInstance();
this.classifier2 = (Classifier) Class.forName(classifier).newInstance();
this.classifier3 = (Classifier) Class.forName(classifier).newInstance();
Instances trainingInstances = Util.getInstances(trainingIns_File);
// 将trainIns_File按照precentage和(1-precentage)的比例切割成labeledIns和unlabeledIns;
int length = trainingInstances.numInstances();
int i = new Double(length * precentage).intValue();
labeledIns = new Instances(trainingInstances, 0);
for (int j = 0; j < i; j++) {
labeledIns.add(trainingInstances.firstInstance());
trainingInstances.delete(0);
}
unlabeledIns = trainingInstances;
testIns = Util.getInstances(testIns_File);
Init();
} catch (Exception e) {
}
}
/**
* Generates a clusterer. Has to initialize all fields of the clusterer that are not being set via
* options.
*
* @param data set of instances serving as training data
* @exception Exception if the clusterer has not been generated successfully
*/
public void buildClusterer(Instances data) throws Exception {
// long start = System.currentTimeMillis();
if (data.checkForStringAttributes()) {
throw new Exception("Can't handle string attributes!");
}
m_ReplaceMissingFilter = new ReplaceMissingValues();
m_ReplaceMissingFilter.setInputFormat(data);
m_instances = Filter.useFilter(data, m_ReplaceMissingFilter);
initMinMax(m_instances);
m_ClusterCentroids = new Instances(m_instances, m_NumClusters);
int n = m_instances.numInstances();
Random r = new Random(m_Seed);
boolean[] selected = new boolean[n];
double[] minDistance = new double[n];
for (int i = 0; i < n; i++) minDistance[i] = Double.MAX_VALUE;
int firstI = r.nextInt(n);
m_ClusterCentroids.add(m_instances.instance(firstI));
selected[firstI] = true;
updateMinDistance(minDistance, selected, m_instances, m_instances.instance(firstI));
if (m_NumClusters > n) m_NumClusters = n;
for (int i = 1; i < m_NumClusters; i++) {
int nextI = farthestAway(minDistance, selected);
m_ClusterCentroids.add(m_instances.instance(nextI));
selected[nextI] = true;
updateMinDistance(minDistance, selected, m_instances, m_instances.instance(nextI));
}
m_instances = new Instances(m_instances, 0);
// long end = System.currentTimeMillis();
// System.out.println("Clustering Time = " + (end-start));
}
// 通过h1,h2分类器学习样本集，将h1,h2分类决策相同的样本放入L中，得到标记集合；
public void updateL(Classifier h1, Classifier h2, Instances L, Instances test) {
int length = unlabeledIns.numInstances();
double value1 = 0.0, value2 = 0.0;
try {
for (int i = 0; i < length; i++) {
value1 = h1.classifyInstance(test.instance(i));
value2 = h2.classifyInstance(test.instance(i));
if (value1 == value2) {
// 当两个分类器做出相同决策时重新标记样本的类别；
test.instance(i).setClassValue(value1);
L.add(test.instance(i));
}
}
} catch (Exception e) {
System.out.println(e);
}
// return false;
}
public void buildClusterer(ArrayList<String> seqDB, double[][] sm) {
seqList = seqDB;
this.setSimMatrix(sm);
Attribute seqString = new Attribute("sequence", (FastVector) null);
FastVector attrInfo = new FastVector();
attrInfo.addElement(seqString);
Instances data = new Instances("data", attrInfo, 0);
for (int i = 0; i < seqList.size(); i++) {
Instance currentInst = new Instance(1);
currentInst.setDataset(data);
currentInst.setValue(0, seqList.get(i));
data.add(currentInst);
}
try {
buildClusterer(data);
} catch (Exception e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
}
/**
* Add new instances to the given set of instances.
*
* @param data given instances
* @param newData set of instances to add to given instances
*/
protected void addInstances(Instances data, Instances newData) {
for (int i = 0; i < newData.numInstances(); i++) data.add(newData.instance(i));
}
/**
* ************************************************** Convert a table to a set of instances, with
* <b>columns</b> representing individual </b>instances</b> and <b>rows</b> representing
* <b>attributes</b> (e.g. as is common with microarray data)
*/
public Instances tableColsToInstances(Table t, String relationName) {
System.err.print("Converting table cols to instances...");
// Set up attributes, which for colInstances will be the rowNames...
FastVector atts = new FastVector();
ArrayList<Boolean> isNominal = new ArrayList<Boolean>();
ArrayList<FastVector> allAttVals = new ArrayList<FastVector>(); // Save values for later...
System.err.print("creating attributes...");
for (int r = 0; r < t.numRows; r++) {
if (rowIsNumeric(t, r)) {
isNominal.add(false);
atts.addElement(new Attribute(t.rowNames[r]));
allAttVals.add(null); // No enumeration of attribute values.
} else {
// It's nominal... determine the range of values and create a nominal attribute...
isNominal.add(true);
FastVector attVals = getRowValues(t, r);
atts.addElement(new Attribute(t.rowNames[r], attVals));
// Save it for later
allAttVals.add(attVals);
}
}
System.err.print("creating instances...");
// Create Instances object..
Instances data = new Instances(relationName, atts, 0);
data.setRelationName(relationName);
/** ***** CREATE INSTANCES ************* */
// Fill the instances with data...
// For each instance...
for (int c = 0; c < t.numCols; c++) {
double[] vals =
new double[data.numAttributes()]; // Even nominal values are stored as double pointers.
// For each attribute fill in the numeric or attributeValue index...
for (int r = 0; r < t.numRows; r++) {
String val = (String) t.matrix.getQuick(r, c);
if (val == "?") vals[r] = Instance.missingValue();
else if (isNominal.get(r)) {
vals[r] = allAttVals.get(r).indexOf(val);
} else {
vals[r] = Double.parseDouble((String) val);
}
}
// Add the a newly minted instance with those attribute values...
data.add(new Instance(1.0, vals));
}
System.err.print("add feature names...");
/** ***** ADD FEATURE NAMES ************* */
// takes basically zero time... all time is in previous 2 chunks.
if (addInstanceNamesAsFeatures) {
Instances newData = new Instances(data);
newData.insertAttributeAt(new Attribute("ID", (FastVector) null), 0);
int attrIdx = newData.attribute("ID").index(); // Paranoid... should be 0
// We save the instanceNames in a list because it's handy later on...
instanceNames = new ArrayList<String>();
for (int c = 0; c < t.colNames.length; c++) {
instanceNames.add(t.colNames[c]);
newData.instance(c).setValue(attrIdx, t.colNames[c]);
}
data = newData;
}
System.err.println("done.");
return (data);
}
/**
* ************************************************** Convert a table to a set of instances, with
* <b>rows</b> representing individual </b>instances</b> and <b>columns</b> representing
* <b>attributes</b>
*/
public Instances tableRowsToNominalInstances(Table t, String relationName) {
System.err.print("Converting table rows to instances...");
// Set up attributes, which for rowInstances will be the colNames...
FastVector atts = new FastVector();
ArrayList<Boolean> isNominal = new ArrayList<Boolean>();
ArrayList<FastVector> allAttVals = new ArrayList<FastVector>(); // Save values for later...
System.err.print("creating attributes...");
for (int c = 0; c < t.numCols; c++) {
// It's nominal... determine the range of values
isNominal.add(true);
FastVector attVals = getColValues(t, c);
atts.addElement(new Attribute(t.colNames[c], attVals));
// Save it for later
allAttVals.add(attVals);
}
System.err.print("creating instances...");
// Create Instances object..
Instances data = new Instances(relationName, atts, 0);
data.setRelationName(relationName);
// Fill the instances with data...
// For each instance...
for (int r = 0; r < t.numRows; r++) {
double[] vals = new double[data.numAttributes()];
// for each attribute
for (int c = 0; c < t.numCols; c++) {
String val = (String) t.matrix.getQuick(r, c);
if (val == "?") vals[c] = Instance.missingValue();
else if (isNominal.get(c)) {
vals[c] = allAttVals.get(c).indexOf(val);
} else {
vals[c] = Double.parseDouble((String) val);
}
}
// Add the a newly minted instance with those attribute values...
data.add(new Instance(1.0, vals));
}
System.err.print("add feature names...");
if (addInstanceNamesAsFeatures) {
Instances newData = new Instances(data);
newData.insertAttributeAt(new Attribute("ID", (FastVector) null), 0);
int attrIdx = newData.attribute("ID").index(); // Paranoid... should be 0
// We save the instanceNames in a list because it's handy later on...
instanceNames = new ArrayList<String>();
for (int r = 0; r < t.rowNames.length; r++) {
instanceNames.add(t.rowNames[r]);
newData.instance(r).setValue(attrIdx, t.rowNames[r]);
}
data = newData;
}
System.err.println("done.");
return (data);
}
/**
* If we know in advance that the table is numeric, can optimize a lot... For example, on 9803 x
* 294 table, TableFileLoader.readNumeric takes 6s compared to 12s for WekaMine readFromTable.
*/
public static Instances readNumeric(String fileName, String relationName, String delimiter)
throws Exception {
int numAttributes = FileUtils.fastCountLines(fileName) - 1; // -1 exclude heading.
String[] attrNames = new String[numAttributes];
// Read the col headings and figure out the number of columns in the table..
BufferedReader reader = new BufferedReader(new FileReader(fileName), 4194304);
String line = reader.readLine();
String[] instanceNames = parseColNames(line, delimiter);
int numInstances = instanceNames.length;
System.err.print("reading " + numAttributes + " x " + numInstances + " table..");
// Create an array to hold the data as we read it in...
double dataArray[][] = new double[numAttributes][numInstances];
// Populate the matrix with values...
String valToken = "";
try {
int rowIdx = 0;
while ((line = reader.readLine()) != null) {
String[] tokens = line.split(delimiter, -1);
attrNames[rowIdx] = tokens[0].trim();
for (int colIdx = 0; colIdx < (tokens.length - 1); colIdx++) {
valToken = tokens[colIdx + 1];
double value;
if (valToken.equals("null")) {
value = Instance.missingValue();
} else if (valToken.equals("?")) {
value = Instance.missingValue();
} else if (valToken.equals("NA")) {
value = Instance.missingValue();
} else if (valToken.equals("")) {
value = Instance.missingValue();
// }else value = DoubleParser.lightningParse(valToken); // faster double parser with
// MANY assumptions
} else value = Double.parseDouble(valToken);
dataArray[rowIdx][colIdx] = value;
}
rowIdx++;
}
} catch (NumberFormatException e) {
System.err.println(e.toString());
System.err.println("Parsing line: " + line);
System.err.println("Parsing token: " + valToken);
}
// Set up attributes, which for colInstances will be the rowNames...
FastVector atts = new FastVector();
for (int a = 0; a < numAttributes; a++) {
atts.addElement(new Attribute(attrNames[a]));
}
// Create Instances object..
Instances data = new Instances(relationName, atts, 0);
data.setRelationName(relationName);
System.err.print("creating instances..");
// System.err.println("DEBUG: numAttributes "+numAttributes);
/** ***** CREATE INSTANCES ************* */
// Fill the instances with data...
// For each instance...
for (int c = 0; c < numInstances; c++) {
double[] vals =
new double[data.numAttributes()]; // Even nominal values are stored as double pointers.
for (int r = 0; r < numAttributes; r++) {
double val = dataArray[r][c];
vals[r] = val;
}
// Add the a newly minted instance with those attribute values...
data.add(new Instance(1.0, vals));
}
// System.err.println("DEBUG: data.numInstances: "+data.numInstances());
// System.err.println("DEBUG: data.numAttributes: "+data.numAttributes());
// System.err.println("DEBUG: data.relationNAme"+data.relationName());
System.err.print("add feature names..");
/** ***** ADD FEATURE NAMES ************* */
// takes basically zero time... all time is in previous 2 chunks.
Instances newData = new Instances(data);
newData.insertAttributeAt(new Attribute("ID", (FastVector) null), 0);
int attrIdx = newData.attribute("ID").index(); // Paranoid... should be 0
for (int c = 0; c < numInstances; c++) {
newData.instance(c).setValue(attrIdx, instanceNames[c]);
}
data = newData;
// System.err.println("DEBUG: data.numInstances: "+data.numInstances());
// System.err.println("DEBUG: data.numAttributes: "+data.numAttributes());
return (data);
}
