instancelist.java

java数据挖掘算法

package shared;
import java.lang.*;
import java.io.*;
import java.util.*;

/** The InstanceList class provides basic functions for ordered lists of
 * instances. Instances may be labelled or unlabelled. Depending on
 * usage, the list may or may not keep counts about various data in the list.
 * These counts are kept in the BagCounters class.                          <P>
 * Assumptions  :                                                           <P>
 * File format follows Quinlan (pp. 81-83) EXCEPT:                          <BR>
 * 1)  , : | \ . do not appear in names                                     <BR>
 * 2) . at end of lists is optional                                         <BR>
 * 3) a word that appears before the labels are enumerated, that is preceded by
 * \ is interpreted as a modifier.  Currently, the only implemented modifier is
 * "weighted", which indicates that the list will be weighted. This means that
 * labels are assumed to be nominal type for read_names().                  <P>
 * Comments     :                                                           <P>
 * Line numbers given are the result of '\n', not wrapping of lines.        <P>
 * "continuous" is not a legal name for the first nominal attribute value; it
 * is reserved to indicate a continuous(RealAttrInfo) attribute.            <P>
 * "discrete" is not a legal name for the first nominal attribute value; it is
 * reserved to indicate a discrete (NominalAttrInfo) attribute but with a
 * dynamic set of values to be specified as they appear in the data file.   <P>
 * "discrete n" is supported, where n is an estimate of the number of values of
 * the attribute.                                                           <P>
 * "nolabel" may be specified as the label field ONLY. If specified, it
 * indicates an unlabelled list.                                            <P>
 *
 * Enhancements :                                                           <P>
 * Cause fatal_error() if read_names() is called by methods other than the
 * constructor or InstanceList.read_names()                                 <P>
 * Extend read_attributes to handle AttrInfo other than NominalAttrInfo and
 * RealAttrInfo.                                                            <P>
 * Expand capability of input function to:                                  <P>
 * 1) allow . in name if not followed by a space                            <P>
 * 2) allow , : | and \ in names if preceded by a backslash                 <P>
 * (this would mimic Quinlan)                                               <P>
 * Use lex to do the lexical analysis of the input file. This will be critical
 * if the syntax becomes more complicated.                                  <P>
 * Ideally impute_unknown_values would  handle both nominal and real values in
 * a single pass.  It should accept an array of operators allowing each
 * attribute to handle unknowns in a different way.  Obvious operators would be:
 * unique_value, mode, mean...                                              <P>
 *
 * @author James Louis 12/08/2000 Ported to Java.
 * @author Alex Kozlov 8/22/96 Added transpose() function.
 * @author Dan Sommerfield 2/22/96 Combined BagSet/InstList/CtrInstList
 * into one class.
 * @author Robert Allen 1/27/95 Modify project() after spreading work to
 * Instance & Schema.
 * @author Richard Long 7/29/93 Initial revision (.h, .c)
 */
public class InstanceList implements Cloneable{
    
    /** The maximum number of attributes allowable in an Instance.**/
    private static int maxAttrVals;
    /** The maximum number of labels allowed for an Instance.**/
    private static int maxLabelVals;
    /** Indicator of whether the Instances in this InstanceList are weighted.**/
    private boolean weighted;
    /** The total weight of all Instances in this InstanceList.**/
    private double totalWeight;
    /** Counts of each classification label found in the Instances stored in
     * this object. **/
    private BagCounters bagCounters;
    /** Schema for the data stored in the file from which data in this object
     * is produced. **/
    private FileSchema fileSchema;
    /** Schema of attriubtes that will actually be used in computation. **/
    private Schema schema;
    /** Indicator for removing all Instances for which there are unknown values
     * on attributes.**/
    private static boolean removeUnknownInstances;
    /** Indicator that Instances that have no weight should be removed from the
     * InstanceList.**/
    private static boolean removeZeroWeights;
    /** The rate at which Instances should have attribute values replaced
     * with unknown values.**/
    private static double corruptToUnknownRate;
    /** The seed for random placement of unknown values.**/
    private static int unknownSeed;
    /** The random number generator used for the placement of unknown values.**/
    private static Random mrandomForUnknowns;
    /** Indicator that this InstanceList has been initialized with Instances**/
    private static boolean initialized;
    /** TRUE if the MineSet program is being used.**/
    private static boolean mineset = false;
    /** The maximum number of warnings that will be logged for unknown labels.**/
    private static int MAX_UNKNOWN_LABEL_WARNING = 10;
    /** The maximum number of warnings that will be logged for Instances with
     * negative weight values.**/
    private static int MAX_NEG_WEIGHT_WARNINGS = 10;
    
    /** The list of Instances.**/
    private LinkedList instances; //list of Instance references
    
    /** LogOptions object containing information for logging purposes.
     */    
    public static LogOptions logOptions = new LogOptions();
    
    /** Constructor.
     * @param file The root name of the file to be loaded into the InstanceList.
     */
    public InstanceList(String file) {
        instances = new LinkedList();
        weighted = false;
        totalWeight = 0;
        bagCounters = null;
        init_max_vals();
        String namesFile = new String(file + Globals.DEFAULT_NAMES_EXT);
        fileSchema = new FileSchema(namesFile);
        schema = fileSchema.create_schema();  //SchemaRC->Schema
        
        fileSchema.display();
        String dataName = file + Globals.DEFAULT_DATA_EXT;
        read_data(dataName,false);
    }
    
    /** Constructor. InstanceList(String, String, String) takes complexity of
     * InstanceList.read_names() + complexity of InstanceList.read_data().
     * @param file The root name of the file to be loaded into the InstanceList.
     * @param namesExtension The file extension for the schema file.
     * @param dataExtension The file extension for the data file.
     */
    public InstanceList( String file,
    String namesExtension,
    String dataExtension) {
        instances = new LinkedList();
        weighted = false;
        totalWeight = 0;
        bagCounters = null;
        init_max_vals();
        String namesFile = new String(file + namesExtension);
        fileSchema = new FileSchema(namesFile);
        schema = fileSchema.create_schema();
        String dataName = file + dataExtension;
        read_data(dataName, false);
    }
    
    /** Constructor.
     * @param catSchema The schema of categories for these data sets.
     * @param file The root name of the file to be loaded into the InstanceList.
     * @param namesExtension The file extension for the schema file.
     * @param testExtension The file extension for the test file.
     */
    public InstanceList(Schema catSchema,
    String file,
    String namesExtension,
    String testExtension) {
        instances = new LinkedList();
        totalWeight = 0;
        try{
            schema = new Schema(catSchema);
        }catch(CloneNotSupportedException e){
            Error.err("InstanceList:constructor(Schema)):clone not"
            +" supported exception caught");}
        fileSchema = null;
        weighted = false;
        bagCounters = null;
        
        init_max_vals();
        String namesFile = file + namesExtension;
        fileSchema = new FileSchema(namesFile);
        read_data(file + testExtension, true);
    }
    
    /** Constructor.
     * @param catSchema The schema of categories for these data sets.
     */
    public InstanceList(Schema catSchema) {
        instances = new LinkedList();
        totalWeight = 0;
        try{
            schema = new Schema(catSchema);
        }catch(CloneNotSupportedException e){
            Error.err("InstanceList:constructor(Schema)):clone not"
            +" supported exception caught");}
        fileSchema = null;
        weighted = false;
        bagCounters = null;
        
        init_max_vals();
    }
    
    /** Constructor.
     * @param catSchema The schema of categories for these data sets.
     * @param names The schema of attributes for these data sets.
     * @param testName The file name for the test file.
     */
    public InstanceList(Schema catSchema,
    FileSchema names,
    String testName) {
        instances = new LinkedList();
        weighted = false;
        totalWeight = 0;
        bagCounters = null;
        init_max_vals();
        fileSchema = new FileSchema(names);
        try{
            schema = new Schema(catSchema);
        }catch(CloneNotSupportedException e){
            Error.err("InstanceList:copyConstructor:clone not"
            +" supported exception caught");}
        read_data(testName, true);
    }
    
    /** Constructor.
     * @param source The InstanceList that is being copied.
     */
    public InstanceList(InstanceList source) {
        boolean preserveCounters = false;
        
        instances = new LinkedList();
        totalWeight = 0;
        try{
            schema = new Schema(source.schema);
        }catch(CloneNotSupportedException e){
            Error.err("InstanceList:copyConstructor:clone not"
            +" supported exception caught");}
        fileSchema = null;
        weighted = source.weighted;
        bagCounters = null;
        
        //weight is accumulated into totalWeight as instances are added.
        init_max_vals();
        ListIterator pix = source.instance_list().listIterator();
        Instance inst = null;
        while(pix.hasNext()) {
            inst = (Instance)pix.next();
            add_instance(inst);
        }
        //If we have a fileSchema, copy it.
        if(source.fileSchema != null)
            fileSchema = new FileSchema(source.fileSchema);
        
        //if we have counters and we want to preserve them, the copy.
        if(source.has_counters() && preserveCounters)
            bagCounters = new BagCounters(source.counters() );
        
        //DBG(OK());
    }
    
    /** Copy constructor.
     * @param source The InstanceList object to be copied.
     * @param preserveCounters TRUE if counters of values should be copied, FALSE otherwise.
     */    
    public InstanceList(InstanceList source, boolean preserveCounters) {
        instances = new LinkedList();
        totalWeight = 0 ; // will get set when instances are added
        try{
            schema = new Schema(source.schema);
        }catch(CloneNotSupportedException e){
            Error.err("InstanceList:copyConstructor:clone not"
            +" supported exception caught");}
        fileSchema = null;
        weighted = source.weighted;
        bagCounters = null;
        
        
        // weight is accumulated into totalWeight as instances are
        // added.
        init_max_vals();
        
        for (ListIterator pix = source.instance_list().listIterator();
        pix.hasNext();)
            add_instance((Instance)pix.next());
        
        // If we have a fileSchema, copy it.
        if(source.fileSchema != null)
            fileSchema = new FileSchema(source.fileSchema);
        
        // If we have counters and we want to preserve them, then copy.
        if(source.has_counters() && preserveCounters)
            bagCounters = new BagCounters(source.counters());
        
        //   if(Globals.DBG(OK();)
    }
    
    /** Build an instance list which is designed to be a test list for some
     * other training set.  The training set must have been built with a
     * FileSchema which will now be used to interpret the test data.
     * @param trainList The training InstanceList that will be used to identify Schema for test data set.
     * @param testName The name of the file containing the test data set.
     */
    public InstanceList(InstanceList trainList,String testName) {
        instances = new LinkedList();
        totalWeight = 0;
        try{
            schema = new Schema(trainList.get_schema());
        }catch(CloneNotSupportedException e){e.printStackTrace();System.exit(1);}