[4] | 1 | /* |
---|
| 2 | * This program is free software; you can redistribute it and/or modify |
---|
| 3 | * it under the terms of the GNU General Public License as published by |
---|
| 4 | * the Free Software Foundation; either version 2 of the License, or |
---|
| 5 | * (at your option) any later version. |
---|
| 6 | * |
---|
| 7 | * This program is distributed in the hope that it will be useful, |
---|
| 8 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
---|
| 9 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
---|
| 10 | * GNU General Public License for more details. |
---|
| 11 | * |
---|
| 12 | * You should have received a copy of the GNU General Public License |
---|
| 13 | * along with this program; if not, write to the Free Software |
---|
| 14 | * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. |
---|
| 15 | */ |
---|
| 16 | |
---|
| 17 | /* |
---|
| 18 | * ConsistencySubsetEval.java |
---|
| 19 | * Copyright (C) 1999 University of Waikato, Hamilton, New Zealand |
---|
| 20 | * |
---|
| 21 | */ |
---|
| 22 | |
---|
| 23 | package weka.attributeSelection; |
---|
| 24 | |
---|
| 25 | import weka.core.Capabilities; |
---|
| 26 | import weka.core.Instance; |
---|
| 27 | import weka.core.Instances; |
---|
| 28 | import weka.core.RevisionHandler; |
---|
| 29 | import weka.core.RevisionUtils; |
---|
| 30 | import weka.core.TechnicalInformation; |
---|
| 31 | import weka.core.TechnicalInformationHandler; |
---|
| 32 | import weka.core.Utils; |
---|
| 33 | import weka.core.Capabilities.Capability; |
---|
| 34 | import weka.core.TechnicalInformation.Field; |
---|
| 35 | import weka.core.TechnicalInformation.Type; |
---|
| 36 | import weka.filters.Filter; |
---|
| 37 | import weka.filters.supervised.attribute.Discretize; |
---|
| 38 | |
---|
| 39 | import java.io.Serializable; |
---|
| 40 | import java.util.BitSet; |
---|
| 41 | import java.util.Enumeration; |
---|
| 42 | import java.util.Hashtable; |
---|
| 43 | |
---|
| 44 | /** |
---|
| 45 | <!-- globalinfo-start --> |
---|
| 46 | * ConsistencySubsetEval :<br/> |
---|
| 47 | * <br/> |
---|
| 48 | * Evaluates the worth of a subset of attributes by the level of consistency in the class values when the training instances are projected onto the subset of attributes. <br/> |
---|
| 49 | * <br/> |
---|
| 50 | * Consistency of any subset can never be lower than that of the full set of attributes, hence the usual practice is to use this subset evaluator in conjunction with a Random or Exhaustive search which looks for the smallest subset with consistency equal to that of the full set of attributes.<br/> |
---|
| 51 | * <br/> |
---|
| 52 | * For more information see:<br/> |
---|
| 53 | * <br/> |
---|
| 54 | * H. Liu, R. Setiono: A probabilistic approach to feature selection - A filter solution. In: 13th International Conference on Machine Learning, 319-327, 1996. |
---|
| 55 | * <p/> |
---|
| 56 | <!-- globalinfo-end --> |
---|
| 57 | * |
---|
| 58 | <!-- technical-bibtex-start --> |
---|
| 59 | * BibTeX: |
---|
| 60 | * <pre> |
---|
| 61 | * @inproceedings{Liu1996, |
---|
| 62 | * author = {H. Liu and R. Setiono}, |
---|
| 63 | * booktitle = {13th International Conference on Machine Learning}, |
---|
| 64 | * pages = {319-327}, |
---|
| 65 | * title = {A probabilistic approach to feature selection - A filter solution}, |
---|
| 66 | * year = {1996} |
---|
| 67 | * } |
---|
| 68 | * </pre> |
---|
| 69 | * <p/> |
---|
| 70 | <!-- technical-bibtex-end --> |
---|
| 71 | * |
---|
| 72 | * @author Mark Hall (mhall@cs.waikato.ac.nz) |
---|
| 73 | * @version $Revision: 5447 $ |
---|
| 74 | * @see Discretize |
---|
| 75 | */ |
---|
| 76 | public class ConsistencySubsetEval |
---|
| 77 | extends ASEvaluation |
---|
| 78 | implements SubsetEvaluator, |
---|
| 79 | TechnicalInformationHandler { |
---|
| 80 | |
---|
| 81 | /** for serialization */ |
---|
| 82 | static final long serialVersionUID = -2880323763295270402L; |
---|
| 83 | |
---|
| 84 | /** training instances */ |
---|
| 85 | private Instances m_trainInstances; |
---|
| 86 | |
---|
| 87 | /** class index */ |
---|
| 88 | private int m_classIndex; |
---|
| 89 | |
---|
| 90 | /** number of attributes in the training data */ |
---|
| 91 | private int m_numAttribs; |
---|
| 92 | |
---|
| 93 | /** number of instances in the training data */ |
---|
| 94 | private int m_numInstances; |
---|
| 95 | |
---|
| 96 | /** Discretise numeric attributes */ |
---|
| 97 | private Discretize m_disTransform; |
---|
| 98 | |
---|
| 99 | /** Hash table for evaluating feature subsets */ |
---|
| 100 | private Hashtable m_table; |
---|
| 101 | |
---|
| 102 | /** |
---|
| 103 | * Class providing keys to the hash table. |
---|
| 104 | */ |
---|
| 105 | public class hashKey |
---|
| 106 | implements Serializable, RevisionHandler { |
---|
| 107 | |
---|
| 108 | /** for serialization */ |
---|
| 109 | static final long serialVersionUID = 6144138512017017408L; |
---|
| 110 | |
---|
| 111 | /** Array of attribute values for an instance */ |
---|
| 112 | private double [] attributes; |
---|
| 113 | |
---|
| 114 | /** True for an index if the corresponding attribute value is missing. */ |
---|
| 115 | private boolean [] missing; |
---|
| 116 | |
---|
| 117 | /** The key */ |
---|
| 118 | private int key; |
---|
| 119 | |
---|
| 120 | /** |
---|
| 121 | * Constructor for a hashKey |
---|
| 122 | * |
---|
| 123 | * @param t an instance from which to generate a key |
---|
| 124 | * @param numAtts the number of attributes |
---|
| 125 | * @throws Exception if something goes wrong |
---|
| 126 | */ |
---|
| 127 | public hashKey(Instance t, int numAtts) throws Exception { |
---|
| 128 | |
---|
| 129 | int i; |
---|
| 130 | int cindex = t.classIndex(); |
---|
| 131 | |
---|
| 132 | key = -999; |
---|
| 133 | attributes = new double [numAtts]; |
---|
| 134 | missing = new boolean [numAtts]; |
---|
| 135 | for (i=0;i<numAtts;i++) { |
---|
| 136 | if (i == cindex) { |
---|
| 137 | missing[i] = true; |
---|
| 138 | } else { |
---|
| 139 | if ((missing[i] = t.isMissing(i)) == false) { |
---|
| 140 | attributes[i] = t.value(i); |
---|
| 141 | } |
---|
| 142 | } |
---|
| 143 | } |
---|
| 144 | } |
---|
| 145 | |
---|
| 146 | /** |
---|
| 147 | * Convert a hash entry to a string |
---|
| 148 | * |
---|
| 149 | * @param t the set of instances |
---|
| 150 | * @param maxColWidth width to make the fields |
---|
| 151 | * @return the hash entry as string |
---|
| 152 | */ |
---|
| 153 | public String toString(Instances t, int maxColWidth) { |
---|
| 154 | |
---|
| 155 | int i; |
---|
| 156 | int cindex = t.classIndex(); |
---|
| 157 | StringBuffer text = new StringBuffer(); |
---|
| 158 | |
---|
| 159 | for (i=0;i<attributes.length;i++) { |
---|
| 160 | if (i != cindex) { |
---|
| 161 | if (missing[i]) { |
---|
| 162 | text.append("?"); |
---|
| 163 | for (int j=0;j<maxColWidth;j++) { |
---|
| 164 | text.append(" "); |
---|
| 165 | } |
---|
| 166 | } else { |
---|
| 167 | String ss = t.attribute(i).value((int)attributes[i]); |
---|
| 168 | StringBuffer sb = new StringBuffer(ss); |
---|
| 169 | |
---|
| 170 | for (int j=0;j < (maxColWidth-ss.length()+1); j++) { |
---|
| 171 | sb.append(" "); |
---|
| 172 | } |
---|
| 173 | text.append(sb); |
---|
| 174 | } |
---|
| 175 | } |
---|
| 176 | } |
---|
| 177 | return text.toString(); |
---|
| 178 | } |
---|
| 179 | |
---|
| 180 | /** |
---|
| 181 | * Constructor for a hashKey |
---|
| 182 | * |
---|
| 183 | * @param t an array of feature values |
---|
| 184 | */ |
---|
| 185 | public hashKey(double [] t) { |
---|
| 186 | |
---|
| 187 | int i; |
---|
| 188 | int l = t.length; |
---|
| 189 | |
---|
| 190 | key = -999; |
---|
| 191 | attributes = new double [l]; |
---|
| 192 | missing = new boolean [l]; |
---|
| 193 | for (i=0;i<l;i++) { |
---|
| 194 | if (t[i] == Double.MAX_VALUE) { |
---|
| 195 | missing[i] = true; |
---|
| 196 | } else { |
---|
| 197 | missing[i] = false; |
---|
| 198 | attributes[i] = t[i]; |
---|
| 199 | } |
---|
| 200 | } |
---|
| 201 | } |
---|
| 202 | |
---|
| 203 | /** |
---|
| 204 | * Calculates a hash code |
---|
| 205 | * |
---|
| 206 | * @return the hash code as an integer |
---|
| 207 | */ |
---|
| 208 | public int hashCode() { |
---|
| 209 | |
---|
| 210 | int hv = 0; |
---|
| 211 | |
---|
| 212 | if (key != -999) |
---|
| 213 | return key; |
---|
| 214 | for (int i=0;i<attributes.length;i++) { |
---|
| 215 | if (missing[i]) { |
---|
| 216 | hv += (i*13); |
---|
| 217 | } else { |
---|
| 218 | hv += (i * 5 * (attributes[i]+1)); |
---|
| 219 | } |
---|
| 220 | } |
---|
| 221 | if (key == -999) { |
---|
| 222 | key = hv; |
---|
| 223 | } |
---|
| 224 | return hv; |
---|
| 225 | } |
---|
| 226 | |
---|
| 227 | /** |
---|
| 228 | * Tests if two instances are equal |
---|
| 229 | * |
---|
| 230 | * @param b a key to compare with |
---|
| 231 | * @return true if the objects are equal |
---|
| 232 | */ |
---|
| 233 | public boolean equals(Object b) { |
---|
| 234 | |
---|
| 235 | if ((b == null) || !(b.getClass().equals(this.getClass()))) { |
---|
| 236 | return false; |
---|
| 237 | } |
---|
| 238 | boolean ok = true; |
---|
| 239 | boolean l; |
---|
| 240 | if (b instanceof hashKey) { |
---|
| 241 | hashKey n = (hashKey)b; |
---|
| 242 | for (int i=0;i<attributes.length;i++) { |
---|
| 243 | l = n.missing[i]; |
---|
| 244 | if (missing[i] || l) { |
---|
| 245 | if ((missing[i] && !l) || (!missing[i] && l)) { |
---|
| 246 | ok = false; |
---|
| 247 | break; |
---|
| 248 | } |
---|
| 249 | } else { |
---|
| 250 | if (attributes[i] != n.attributes[i]) { |
---|
| 251 | ok = false; |
---|
| 252 | break; |
---|
| 253 | } |
---|
| 254 | } |
---|
| 255 | } |
---|
| 256 | } else { |
---|
| 257 | return false; |
---|
| 258 | } |
---|
| 259 | return ok; |
---|
| 260 | } |
---|
| 261 | |
---|
| 262 | /** |
---|
| 263 | * Prints the hash code |
---|
| 264 | */ |
---|
| 265 | public void print_hash_code() { |
---|
| 266 | |
---|
| 267 | System.out.println("Hash val: "+hashCode()); |
---|
| 268 | } |
---|
| 269 | |
---|
| 270 | /** |
---|
| 271 | * Returns the revision string. |
---|
| 272 | * |
---|
| 273 | * @return the revision |
---|
| 274 | */ |
---|
| 275 | public String getRevision() { |
---|
| 276 | return RevisionUtils.extract("$Revision: 5447 $"); |
---|
| 277 | } |
---|
| 278 | } |
---|
| 279 | |
---|
| 280 | /** |
---|
| 281 | * Returns a string describing this search method |
---|
| 282 | * @return a description of the search suitable for |
---|
| 283 | * displaying in the explorer/experimenter gui |
---|
| 284 | */ |
---|
| 285 | public String globalInfo() { |
---|
| 286 | return "ConsistencySubsetEval :\n\nEvaluates the worth of a subset of " |
---|
| 287 | +"attributes by the level of consistency in the class values when the " |
---|
| 288 | +"training instances are projected onto the subset of attributes. " |
---|
| 289 | +"\n\nConsistency of any subset can never be lower than that of the " |
---|
| 290 | +"full set of attributes, hence the usual practice is to use this " |
---|
| 291 | +"subset evaluator in conjunction with a Random or Exhaustive search " |
---|
| 292 | +"which looks for the smallest subset with consistency equal to that " |
---|
| 293 | +"of the full set of attributes.\n\n" |
---|
| 294 | + "For more information see:\n\n" |
---|
| 295 | + getTechnicalInformation().toString(); |
---|
| 296 | } |
---|
| 297 | |
---|
| 298 | /** |
---|
| 299 | * Returns an instance of a TechnicalInformation object, containing |
---|
| 300 | * detailed information about the technical background of this class, |
---|
| 301 | * e.g., paper reference or book this class is based on. |
---|
| 302 | * |
---|
| 303 | * @return the technical information about this class |
---|
| 304 | */ |
---|
| 305 | public TechnicalInformation getTechnicalInformation() { |
---|
| 306 | TechnicalInformation result; |
---|
| 307 | |
---|
| 308 | result = new TechnicalInformation(Type.INPROCEEDINGS); |
---|
| 309 | result.setValue(Field.AUTHOR, "H. Liu and R. Setiono"); |
---|
| 310 | result.setValue(Field.TITLE, "A probabilistic approach to feature selection - A filter solution"); |
---|
| 311 | result.setValue(Field.BOOKTITLE, "13th International Conference on Machine Learning"); |
---|
| 312 | result.setValue(Field.YEAR, "1996"); |
---|
| 313 | result.setValue(Field.PAGES, "319-327"); |
---|
| 314 | |
---|
| 315 | return result; |
---|
| 316 | } |
---|
| 317 | |
---|
| 318 | /** |
---|
| 319 | * Constructor. Calls restOptions to set default options |
---|
| 320 | **/ |
---|
| 321 | public ConsistencySubsetEval () { |
---|
| 322 | resetOptions(); |
---|
| 323 | } |
---|
| 324 | |
---|
| 325 | /** |
---|
| 326 | * reset to defaults |
---|
| 327 | */ |
---|
| 328 | private void resetOptions () { |
---|
| 329 | m_trainInstances = null; |
---|
| 330 | } |
---|
| 331 | |
---|
| 332 | /** |
---|
| 333 | * Returns the capabilities of this evaluator. |
---|
| 334 | * |
---|
| 335 | * @return the capabilities of this evaluator |
---|
| 336 | * @see Capabilities |
---|
| 337 | */ |
---|
| 338 | public Capabilities getCapabilities() { |
---|
| 339 | Capabilities result = super.getCapabilities(); |
---|
| 340 | result.disableAll(); |
---|
| 341 | |
---|
| 342 | // attributes |
---|
| 343 | result.enable(Capability.NOMINAL_ATTRIBUTES); |
---|
| 344 | result.enable(Capability.NUMERIC_ATTRIBUTES); |
---|
| 345 | result.enable(Capability.DATE_ATTRIBUTES); |
---|
| 346 | result.enable(Capability.MISSING_VALUES); |
---|
| 347 | |
---|
| 348 | // class |
---|
| 349 | result.enable(Capability.NOMINAL_CLASS); |
---|
| 350 | result.enable(Capability.MISSING_CLASS_VALUES); |
---|
| 351 | |
---|
| 352 | return result; |
---|
| 353 | } |
---|
| 354 | |
---|
| 355 | /** |
---|
| 356 | * Generates a attribute evaluator. Has to initialize all fields of the |
---|
| 357 | * evaluator that are not being set via options. |
---|
| 358 | * |
---|
| 359 | * @param data set of instances serving as training data |
---|
| 360 | * @throws Exception if the evaluator has not been |
---|
| 361 | * generated successfully |
---|
| 362 | */ |
---|
| 363 | public void buildEvaluator (Instances data) throws Exception { |
---|
| 364 | |
---|
| 365 | // can evaluator handle data? |
---|
| 366 | getCapabilities().testWithFail(data); |
---|
| 367 | |
---|
| 368 | m_trainInstances = new Instances(data); |
---|
| 369 | m_trainInstances.deleteWithMissingClass(); |
---|
| 370 | m_classIndex = m_trainInstances.classIndex(); |
---|
| 371 | m_numAttribs = m_trainInstances.numAttributes(); |
---|
| 372 | m_numInstances = m_trainInstances.numInstances(); |
---|
| 373 | |
---|
| 374 | m_disTransform = new Discretize(); |
---|
| 375 | m_disTransform.setUseBetterEncoding(true); |
---|
| 376 | m_disTransform.setInputFormat(m_trainInstances); |
---|
| 377 | m_trainInstances = Filter.useFilter(m_trainInstances, m_disTransform); |
---|
| 378 | } |
---|
| 379 | |
---|
| 380 | /** |
---|
| 381 | * Evaluates a subset of attributes |
---|
| 382 | * |
---|
| 383 | * @param subset a bitset representing the attribute subset to be |
---|
| 384 | * evaluated |
---|
| 385 | * @throws Exception if the subset could not be evaluated |
---|
| 386 | */ |
---|
| 387 | public double evaluateSubset (BitSet subset) throws Exception { |
---|
| 388 | int [] fs; |
---|
| 389 | int i; |
---|
| 390 | int count = 0; |
---|
| 391 | |
---|
| 392 | for (i=0;i<m_numAttribs;i++) { |
---|
| 393 | if (subset.get(i)) { |
---|
| 394 | count++; |
---|
| 395 | } |
---|
| 396 | } |
---|
| 397 | |
---|
| 398 | double [] instArray = new double[count]; |
---|
| 399 | int index = 0; |
---|
| 400 | fs = new int[count]; |
---|
| 401 | for (i=0;i<m_numAttribs;i++) { |
---|
| 402 | if (subset.get(i)) { |
---|
| 403 | fs[index++] = i; |
---|
| 404 | } |
---|
| 405 | } |
---|
| 406 | |
---|
| 407 | // create new hash table |
---|
| 408 | m_table = new Hashtable((int)(m_numInstances * 1.5)); |
---|
| 409 | |
---|
| 410 | for (i=0;i<m_numInstances;i++) { |
---|
| 411 | Instance inst = m_trainInstances.instance(i); |
---|
| 412 | for (int j=0;j<fs.length;j++) { |
---|
| 413 | if (fs[j] == m_classIndex) { |
---|
| 414 | throw new Exception("A subset should not contain the class!"); |
---|
| 415 | } |
---|
| 416 | if (inst.isMissing(fs[j])) { |
---|
| 417 | instArray[j] = Double.MAX_VALUE; |
---|
| 418 | } else { |
---|
| 419 | instArray[j] = inst.value(fs[j]); |
---|
| 420 | } |
---|
| 421 | } |
---|
| 422 | insertIntoTable(inst, instArray); |
---|
| 423 | } |
---|
| 424 | |
---|
| 425 | return consistencyCount(); |
---|
| 426 | } |
---|
| 427 | |
---|
| 428 | /** |
---|
| 429 | * calculates the level of consistency in a dataset using a subset of |
---|
| 430 | * features. The consistency of a hash table entry is the total number |
---|
| 431 | * of instances hashed to that location minus the number of instances in |
---|
| 432 | * the largest class hashed to that location. The total consistency is |
---|
| 433 | * 1.0 minus the sum of the individual consistencies divided by the |
---|
| 434 | * total number of instances. |
---|
| 435 | * @return the consistency of the hash table as a value between 0 and 1. |
---|
| 436 | */ |
---|
| 437 | private double consistencyCount() { |
---|
| 438 | Enumeration e = m_table.keys(); |
---|
| 439 | double [] classDist; |
---|
| 440 | double count = 0.0; |
---|
| 441 | |
---|
| 442 | while (e.hasMoreElements()) { |
---|
| 443 | hashKey tt = (hashKey)e.nextElement(); |
---|
| 444 | classDist = (double []) m_table.get(tt); |
---|
| 445 | count += Utils.sum(classDist); |
---|
| 446 | int max = Utils.maxIndex(classDist); |
---|
| 447 | count -= classDist[max]; |
---|
| 448 | } |
---|
| 449 | |
---|
| 450 | count /= (double)m_numInstances; |
---|
| 451 | return (1.0 - count); |
---|
| 452 | } |
---|
| 453 | |
---|
| 454 | /** |
---|
| 455 | * Inserts an instance into the hash table |
---|
| 456 | * |
---|
| 457 | * @param inst instance to be inserted |
---|
| 458 | * @param instA the instance to be inserted as an array of attribute |
---|
| 459 | * values. |
---|
| 460 | * @throws Exception if the instance can't be inserted |
---|
| 461 | */ |
---|
| 462 | private void insertIntoTable(Instance inst, double [] instA) |
---|
| 463 | throws Exception { |
---|
| 464 | |
---|
| 465 | double [] tempClassDist2; |
---|
| 466 | double [] newDist; |
---|
| 467 | hashKey thekey; |
---|
| 468 | |
---|
| 469 | thekey = new hashKey(instA); |
---|
| 470 | |
---|
| 471 | // see if this one is already in the table |
---|
| 472 | tempClassDist2 = (double []) m_table.get(thekey); |
---|
| 473 | if (tempClassDist2 == null) { |
---|
| 474 | newDist = new double [m_trainInstances.classAttribute().numValues()]; |
---|
| 475 | newDist[(int)inst.classValue()] = inst.weight(); |
---|
| 476 | |
---|
| 477 | // add to the table |
---|
| 478 | m_table.put(thekey, newDist); |
---|
| 479 | } else { |
---|
| 480 | // update the distribution for this instance |
---|
| 481 | tempClassDist2[(int)inst.classValue()]+=inst.weight(); |
---|
| 482 | |
---|
| 483 | // update the table |
---|
| 484 | m_table.put(thekey, tempClassDist2); |
---|
| 485 | } |
---|
| 486 | } |
---|
| 487 | |
---|
| 488 | /** |
---|
| 489 | * returns a description of the evaluator |
---|
| 490 | * @return a description of the evaluator as a String. |
---|
| 491 | */ |
---|
| 492 | public String toString() { |
---|
| 493 | StringBuffer text = new StringBuffer(); |
---|
| 494 | |
---|
| 495 | if (m_trainInstances == null) { |
---|
| 496 | text.append("\tConsistency subset evaluator has not been built yet\n"); |
---|
| 497 | } |
---|
| 498 | else { |
---|
| 499 | text.append("\tConsistency Subset Evaluator\n"); |
---|
| 500 | } |
---|
| 501 | |
---|
| 502 | return text.toString(); |
---|
| 503 | } |
---|
| 504 | |
---|
| 505 | /** |
---|
| 506 | * Returns the revision string. |
---|
| 507 | * |
---|
| 508 | * @return the revision |
---|
| 509 | */ |
---|
| 510 | public String getRevision() { |
---|
| 511 | return RevisionUtils.extract("$Revision: 5447 $"); |
---|
| 512 | } |
---|
| 513 | |
---|
| 514 | /** |
---|
| 515 | * Main method for testing this class. |
---|
| 516 | * |
---|
| 517 | * @param args the options |
---|
| 518 | */ |
---|
| 519 | public static void main (String[] args) { |
---|
| 520 | runEvaluator(new ConsistencySubsetEval(), args); |
---|
| 521 | } |
---|
| 522 | } |
---|