instances.java

Java 编写的多种数据挖掘算法 包括聚类、分类、预处理等

      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 
   * @throws IllegalArgumentException if the number of folds is less than 2
   * or greater than the number of instances.
   */
  //@ requires 2 <= numFolds && numFolds < numInstances();
  //@ requires 0 <= numFold && numFold < numFolds;
  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;
  }

  /**
   * Creates the training set for one fold of a cross-validation 
   * on the dataset. The data is subsequently randomized based
   * on the given random number generator.
   *
   * @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, ...
   * @param random the random number generator
   * @return the training set 
   * @throws IllegalArgumentException if the number of folds is less than 2
   * or greater than the number of instances.
   */
  //@ requires 2 <= numFolds && numFolds < numInstances();
  //@ requires 0 <= numFold && numFold < numFolds;
  public Instances trainCV(int numFolds, int numFold, Random random) {

    Instances train = trainCV(numFolds, numFold);
    train.randomize(random);
    return train;
  }

  /**
   * Computes the variance for a numeric attribute.
   *
   * @param attIndex the numeric attribute (index starts with 0)
   * @return the variance if the attribute is numeric
   * @throws IllegalArgumentException if the attribute is not numeric
   */
  public /*@pure@*/ 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 (sumOfWeights <= 1) {
      return 0;
    }
    double result = (sumSquared - (sum * sum / sumOfWeights)) / 
      (sumOfWeights - 1);

    // We don't like negative variance
    if (result < 0) {
      return 0;
    } else {
      return result;
    }
  }

  /**
   * Computes the variance for a numeric attribute.
   *
   * @param att the numeric attribute
   * @return the variance if the attribute is numeric
   * @throws IllegalArgumentException if the attribute is not numeric
   */
  public /*@pure@*/ 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 (index starts with 0)
   * @return an AttributeStats object with it's fields calculated.
   */
  //@ requires 0 <= index && index < numAttributes();
  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 (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. 
   */
  //@ requires 0 <= index && index < numAttributes();
  public /*@pure@*/ 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.DATE:
	result.append(Utils.padLeft("Dat", 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;
      case Attribute.RELATIONAL:
	result.append(Utils.padLeft("Rel", 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
   * @throws 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
   * @throws 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");
	}
        switch (attribute(m_IndicesBuffer[numValues]).type()) {
          case Attribute.NOMINAL:
            // 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;
            break;
	case Attribute.NUMERIC:
	  // Check if value is really a number.
	  try{
	    m_ValueBuffer[numValues] = Double.valueOf(tokenizer.sval).
	      doubleValue();
	  } catch (NumberFormatException e) {
	    errms(tokenizer,"number expected");
	  }
          break;
	case Attribute.STRING:
	  m_ValueBuffer[numValues] = 
	    attribute(m_IndicesBuffer[numValues]).addStringValue(tokenizer.sval);
          break;
        case Attribute.DATE:
          try {
            m_ValueBuffer[numValues] = 
              attribute(m_IndicesBuffer[numValues]).parseDate(tokenizer.sval);
          } catch (ParseException e) {
            errms(tokenizer,"unparseable date: " + tokenizer.sval);
          }
          break;
        case Attribute.RELATIONAL:
          StringReader reader = new StringReader(tokenizer.sval);
          StreamTokenizer innerTokenizer = new StreamTokenizer(reader);
          initTokenizer(innerTokenizer);
          Instances data = new Instances(attribute(m_IndicesBuffer[numValues]).relation(), 100);

          // Allocate buffers in case sparse instances have to be read
          data.m_ValueBuffer = new double[data.numAttributes()];
          data.m_IndicesBuffer = new int[data.numAttributes()];

          while (data.getInstance(innerTokenizer, true)) {};
          data.compactify();
          m_ValueBuffer[numValues] = attribute(m_IndicesBuffer[numValues]).addRelation(data);
          break;
        default:
          errms(tokenizer,"unknown attribute type in column " + m_IndicesBuffer[numValues]);
	}
      }
      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;
  }