[29] | 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 | * ScatterSearchV1.java |
---|
| 19 | * Copyright (C) 2008 Adrian Pino |
---|
| 20 | * Copyright (C) 2008 University of Waikato, Hamilton, NZ |
---|
| 21 | * |
---|
| 22 | */ |
---|
| 23 | |
---|
| 24 | package weka.attributeSelection; |
---|
| 25 | |
---|
| 26 | |
---|
| 27 | import java.io.Serializable; |
---|
| 28 | import java.util.ArrayList; |
---|
| 29 | import java.util.BitSet; |
---|
| 30 | import java.util.Enumeration; |
---|
| 31 | import java.util.List; |
---|
| 32 | import java.util.Random; |
---|
| 33 | import java.util.Vector; |
---|
| 34 | |
---|
| 35 | import weka.core.Instances; |
---|
| 36 | import weka.core.Option; |
---|
| 37 | import weka.core.OptionHandler; |
---|
| 38 | import weka.core.RevisionUtils; |
---|
| 39 | import weka.core.SelectedTag; |
---|
| 40 | import weka.core.Tag; |
---|
| 41 | import weka.core.TechnicalInformation; |
---|
| 42 | import weka.core.TechnicalInformationHandler; |
---|
| 43 | import weka.core.Utils; |
---|
| 44 | import weka.core.TechnicalInformation.Field; |
---|
| 45 | import weka.core.TechnicalInformation.Type; |
---|
| 46 | |
---|
| 47 | |
---|
| 48 | /** |
---|
| 49 | * Class for performing the Sequential Scatter Search. <p> |
---|
| 50 | * |
---|
| 51 | <!-- globalinfo-start --> |
---|
| 52 | * Scatter Search :<br/> |
---|
| 53 | * <br/> |
---|
| 54 | * Performs an Scatter Search through the space of attribute subsets. Start with a population of many significants and diverses subset stops when the result is higher than a given treshold or there's not more improvement<br/> |
---|
| 55 | * For more information see:<br/> |
---|
| 56 | * <br/> |
---|
| 57 | * Felix Garcia Lopez (2004). Solving feature subset selection problem by a Parallel Scatter Search. Elsevier. |
---|
| 58 | * <p/> |
---|
| 59 | <!-- globalinfo-end --> |
---|
| 60 | * |
---|
| 61 | <!-- options-start --> |
---|
| 62 | * Valid options are: <p/> |
---|
| 63 | * |
---|
| 64 | * <pre> -Z <num> |
---|
| 65 | * Specify the number of subsets to generate |
---|
| 66 | * in the initial population..</pre> |
---|
| 67 | * |
---|
| 68 | * <pre> -T <threshold> |
---|
| 69 | * Specify the treshold used for considering when a subset is significant.</pre> |
---|
| 70 | * |
---|
| 71 | * <pre> -R <0 = greedy combination | 1 = reduced greedy combination > |
---|
| 72 | * Specify the kind of combiantion |
---|
| 73 | * for using it in the combination method.</pre> |
---|
| 74 | * |
---|
| 75 | * <pre> -S <seed> |
---|
| 76 | * Set the random number seed. |
---|
| 77 | * (default = 1)</pre> |
---|
| 78 | * |
---|
| 79 | * <pre> -D |
---|
| 80 | * Verbose output for monitoring the search.</pre> |
---|
| 81 | * |
---|
| 82 | <!-- options-end --> |
---|
| 83 | * |
---|
| 84 | <!-- technical-bibtex-start --> |
---|
| 85 | * BibTeX: |
---|
| 86 | * <pre> |
---|
| 87 | * @book{Lopez2004, |
---|
| 88 | * author = {Felix Garcia Lopez}, |
---|
| 89 | * month = {October}, |
---|
| 90 | * publisher = {Elsevier}, |
---|
| 91 | * title = {Solving feature subset selection problem by a Parallel Scatter Search}, |
---|
| 92 | * year = {2004}, |
---|
| 93 | * language = {English} |
---|
| 94 | * } |
---|
| 95 | * </pre> |
---|
| 96 | * <p/> |
---|
| 97 | <!-- technical-bibtex-end --> |
---|
| 98 | * |
---|
| 99 | * from the Book: Solving feature subset selection problem by a Parallel Scatter Search, Felix Garcia Lopez. |
---|
| 100 | * @author Adrian Pino (apinoa@facinf.uho.edu.cu) |
---|
| 101 | * @version $Revision: 4977 $ |
---|
| 102 | * |
---|
| 103 | */ |
---|
| 104 | |
---|
| 105 | public class ScatterSearchV1 extends ASSearch |
---|
| 106 | implements OptionHandler, TechnicalInformationHandler { |
---|
| 107 | |
---|
| 108 | /** for serialization */ |
---|
| 109 | static final long serialVersionUID = -8512041420388121326L; |
---|
| 110 | |
---|
| 111 | /** number of attributes in the data */ |
---|
| 112 | private int m_numAttribs; |
---|
| 113 | |
---|
| 114 | /** holds the class index */ |
---|
| 115 | private int m_classIndex; |
---|
| 116 | |
---|
| 117 | /** holds the treshhold that delimits the good attributes */ |
---|
| 118 | private double m_treshold; |
---|
| 119 | |
---|
| 120 | /** the initial threshold */ |
---|
| 121 | private double m_initialThreshold; |
---|
| 122 | |
---|
| 123 | /** the kind of comination betwen parents ((0)greedy combination/(1)reduced greedy combination)*/ |
---|
| 124 | int m_typeOfCombination; |
---|
| 125 | |
---|
| 126 | /** random number generation */ |
---|
| 127 | private Random m_random; |
---|
| 128 | |
---|
| 129 | /** seed for random number generation */ |
---|
| 130 | private int m_seed; |
---|
| 131 | |
---|
| 132 | /** verbose output for monitoring the search and debugging */ |
---|
| 133 | private boolean m_debug = false; |
---|
| 134 | |
---|
| 135 | /** holds a report of the search */ |
---|
| 136 | private StringBuffer m_InformationReports; |
---|
| 137 | |
---|
| 138 | /** total number of subsets evaluated during a search */ |
---|
| 139 | private int m_totalEvals; |
---|
| 140 | |
---|
| 141 | /** holds the merit of the best subset found */ |
---|
| 142 | protected double m_bestMerit; |
---|
| 143 | |
---|
| 144 | /** time for procesing the search method */ |
---|
| 145 | private long m_processinTime; |
---|
| 146 | |
---|
| 147 | /** holds the Initial Population of Subsets*/ |
---|
| 148 | private List<Subset> m_population; |
---|
| 149 | |
---|
| 150 | /** holds the population size*/ |
---|
| 151 | private int m_popSize; |
---|
| 152 | |
---|
| 153 | /** holds the user selected initial population size */ |
---|
| 154 | private int m_initialPopSize; |
---|
| 155 | |
---|
| 156 | /** if no initial user pop size, then this holds the initial |
---|
| 157 | * pop size calculated from the number of attributes in the data |
---|
| 158 | * (for use in the toString() method) |
---|
| 159 | */ |
---|
| 160 | private int m_calculatedInitialPopSize; |
---|
| 161 | |
---|
| 162 | /** holds the subsets most significants and diverses |
---|
| 163 | * of the population (ReferenceSet). |
---|
| 164 | * |
---|
| 165 | * (transient because the subList() method returns |
---|
| 166 | * a non serializable Object). |
---|
| 167 | */ |
---|
| 168 | private transient List<Subset> m_ReferenceSet; |
---|
| 169 | |
---|
| 170 | /** holds the greedy combination(reduced or not) of all the subsets of the ReferenceSet*/ |
---|
| 171 | private transient List<Subset> m_parentsCombination; |
---|
| 172 | |
---|
| 173 | /**holds the attributes ranked*/ |
---|
| 174 | private List<Subset> m_attributeRanking; |
---|
| 175 | |
---|
| 176 | /**Evaluator used to know the significance of a subset (for guiding the search)*/ |
---|
| 177 | private SubsetEvaluator ASEvaluator =null; |
---|
| 178 | |
---|
| 179 | |
---|
| 180 | /** kind of combination */ |
---|
| 181 | protected static final int COMBINATION_NOT_REDUCED = 0; |
---|
| 182 | protected static final int COMBINATION_REDUCED = 1; ; |
---|
| 183 | public static final Tag [] TAGS_SELECTION = { |
---|
| 184 | new Tag(COMBINATION_NOT_REDUCED, "Greedy Combination"), |
---|
| 185 | new Tag(COMBINATION_REDUCED, "Reduced Greedy Combination") |
---|
| 186 | }; |
---|
| 187 | |
---|
| 188 | /** |
---|
| 189 | * Returns a string describing this search method |
---|
| 190 | * @return a description of the search suitable for |
---|
| 191 | * displaying in the explorer/experimenter gui |
---|
| 192 | */ |
---|
| 193 | public String globalInfo() { |
---|
| 194 | return "Scatter Search :\n\nPerforms an Scatter Search " |
---|
| 195 | +"through " |
---|
| 196 | +"the space of attribute subsets. Start with a population of many significants and diverses subset " |
---|
| 197 | +" stops when the result is higher than a given treshold or there's not more improvement\n" |
---|
| 198 | + "For more information see:\n\n" |
---|
| 199 | + getTechnicalInformation().toString(); |
---|
| 200 | } |
---|
| 201 | |
---|
| 202 | /** |
---|
| 203 | * Returns an instance of a TechnicalInformation object, containing |
---|
| 204 | * detailed information about the technical background of this class, |
---|
| 205 | * e.g., paper reference or book this class is based on. |
---|
| 206 | * |
---|
| 207 | * @return the technical information about this class |
---|
| 208 | */ |
---|
| 209 | public TechnicalInformation getTechnicalInformation() { |
---|
| 210 | TechnicalInformation result; |
---|
| 211 | |
---|
| 212 | result = new TechnicalInformation(Type.BOOK); |
---|
| 213 | result.setValue(Field.AUTHOR, "Felix Garcia Lopez"); |
---|
| 214 | result.setValue(Field.MONTH, "October"); |
---|
| 215 | result.setValue(Field.YEAR, "2004"); |
---|
| 216 | result.setValue(Field.TITLE, "Solving feature subset selection problem by a Parallel Scatter Search"); |
---|
| 217 | result.setValue(Field.PUBLISHER, "Elsevier"); |
---|
| 218 | result.setValue(Field.LANGUAGE, "English"); |
---|
| 219 | |
---|
| 220 | return result; |
---|
| 221 | } |
---|
| 222 | |
---|
| 223 | /** |
---|
| 224 | * Returns the revision string. |
---|
| 225 | * |
---|
| 226 | * @return the revision |
---|
| 227 | */ |
---|
| 228 | public String getRevision() { |
---|
| 229 | return RevisionUtils.extract("$Revision: 1.0$"); |
---|
| 230 | } |
---|
| 231 | |
---|
| 232 | public ScatterSearchV1 () { |
---|
| 233 | resetOptions(); |
---|
| 234 | } |
---|
| 235 | |
---|
| 236 | /** |
---|
| 237 | * Returns the tip text for this property |
---|
| 238 | * @return tip text for this property suitable for |
---|
| 239 | * displaying in the explorer/experimenter gui |
---|
| 240 | */ |
---|
| 241 | public String tresholdTipText() { |
---|
| 242 | return "Set the treshold that subsets most overcome to be considered as significants"; |
---|
| 243 | } |
---|
| 244 | |
---|
| 245 | /** |
---|
| 246 | * Set the treshold |
---|
| 247 | * |
---|
| 248 | * @param treshold for identifyng significant subsets |
---|
| 249 | */ |
---|
| 250 | public void setTreshold (double treshold) { |
---|
| 251 | m_initialThreshold = treshold; |
---|
| 252 | } |
---|
| 253 | |
---|
| 254 | /** |
---|
| 255 | * Get the treshold |
---|
| 256 | * |
---|
| 257 | * @return the treshold that subsets most overcome to be considered as significants |
---|
| 258 | */ |
---|
| 259 | public double getTreshold () { |
---|
| 260 | return m_initialThreshold; |
---|
| 261 | } |
---|
| 262 | |
---|
| 263 | |
---|
| 264 | /** |
---|
| 265 | * Returns the tip text for this property |
---|
| 266 | * @return tip text for this property suitable for |
---|
| 267 | * displaying in the explorer/experimenter gui |
---|
| 268 | */ |
---|
| 269 | public String populationSizeTipText() { |
---|
| 270 | return "Set the number of subset to generate in the initial Population"; |
---|
| 271 | } |
---|
| 272 | |
---|
| 273 | /** |
---|
| 274 | * Set the population size |
---|
| 275 | * |
---|
| 276 | * @param size the number of subset in the initial population |
---|
| 277 | */ |
---|
| 278 | public void setPopulationSize (int size) { |
---|
| 279 | m_initialPopSize = size; |
---|
| 280 | } |
---|
| 281 | |
---|
| 282 | /** |
---|
| 283 | * Get the population size |
---|
| 284 | * |
---|
| 285 | * @return the number of subsets to generate in the initial population |
---|
| 286 | */ |
---|
| 287 | public int getPopulationSize () { |
---|
| 288 | return m_initialPopSize; |
---|
| 289 | } |
---|
| 290 | |
---|
| 291 | |
---|
| 292 | /** |
---|
| 293 | * Returns the tip text for this property |
---|
| 294 | * @return tip text for this property suitable for |
---|
| 295 | * displaying in the explorer/experimenter gui |
---|
| 296 | */ |
---|
| 297 | public String combinationTipText() { |
---|
| 298 | return "Set the kind of combination for using it to combine ReferenceSet subsets."; |
---|
| 299 | } |
---|
| 300 | |
---|
| 301 | /** |
---|
| 302 | * Set the kind of combination |
---|
| 303 | * |
---|
| 304 | * @param c the kind of combination of the search |
---|
| 305 | */ |
---|
| 306 | public void setCombination (SelectedTag c) { |
---|
| 307 | if (c.getTags() == TAGS_SELECTION) { |
---|
| 308 | m_typeOfCombination = c.getSelectedTag().getID(); |
---|
| 309 | } |
---|
| 310 | } |
---|
| 311 | |
---|
| 312 | /** |
---|
| 313 | * Get the combination |
---|
| 314 | * |
---|
| 315 | * @return the kind of combination used in the Combination method |
---|
| 316 | */ |
---|
| 317 | public SelectedTag getCombination () { |
---|
| 318 | return new SelectedTag(m_typeOfCombination, TAGS_SELECTION); |
---|
| 319 | } |
---|
| 320 | |
---|
| 321 | /** |
---|
| 322 | * Returns the tip text for this property |
---|
| 323 | * @return tip text for this property suitable for |
---|
| 324 | * displaying in the explorer/experimenter gui |
---|
| 325 | */ |
---|
| 326 | public String seedTipText() { |
---|
| 327 | return "Set the random seed."; |
---|
| 328 | } |
---|
| 329 | |
---|
| 330 | /** |
---|
| 331 | * set the seed for random number generation |
---|
| 332 | * @param s seed value |
---|
| 333 | */ |
---|
| 334 | public void setSeed(int s) { |
---|
| 335 | m_seed = s; |
---|
| 336 | } |
---|
| 337 | |
---|
| 338 | /** |
---|
| 339 | * get the value of the random number generator's seed |
---|
| 340 | * @return the seed for random number generation |
---|
| 341 | */ |
---|
| 342 | public int getSeed() { |
---|
| 343 | return m_seed; |
---|
| 344 | } |
---|
| 345 | |
---|
| 346 | /** |
---|
| 347 | * Returns the tip text for this property |
---|
| 348 | * @return tip text for this property suitable for |
---|
| 349 | * displaying in the explorer/experimenter gui |
---|
| 350 | */ |
---|
| 351 | public String debugTipText() { |
---|
| 352 | return "Turn on verbose output for monitoring the search's progress."; |
---|
| 353 | } |
---|
| 354 | |
---|
| 355 | /** |
---|
| 356 | * Set whether verbose output should be generated. |
---|
| 357 | * @param d true if output is to be verbose. |
---|
| 358 | */ |
---|
| 359 | public void setDebug(boolean d) { |
---|
| 360 | m_debug = d; |
---|
| 361 | } |
---|
| 362 | |
---|
| 363 | /** |
---|
| 364 | * Get whether output is to be verbose |
---|
| 365 | * @return true if output will be verbose |
---|
| 366 | */ |
---|
| 367 | public boolean getDebug() { |
---|
| 368 | return m_debug; |
---|
| 369 | } |
---|
| 370 | |
---|
| 371 | /** |
---|
| 372 | * Returns an enumeration describing the available options. |
---|
| 373 | * @return an enumeration of all the available options. |
---|
| 374 | **/ |
---|
| 375 | public Enumeration listOptions () { |
---|
| 376 | Vector newVector = new Vector(6); |
---|
| 377 | |
---|
| 378 | newVector.addElement(new Option("\tSpecify the number of subsets to generate " |
---|
| 379 | + "\n\tin the initial population.." |
---|
| 380 | ,"Z",1 |
---|
| 381 | , "-Z <num>")); |
---|
| 382 | newVector.addElement(new Option("\tSpecify the treshold used for considering when a subset is significant." |
---|
| 383 | , "T", 1 |
---|
| 384 | , "-T <threshold>")); |
---|
| 385 | newVector.addElement(new Option("\tSpecify the kind of combiantion " |
---|
| 386 | + "\n\tfor using it in the combination method." |
---|
| 387 | , "R", 1, "-R <0 = greedy combination | 1 = reduced greedy combination >")); |
---|
| 388 | newVector.addElement(new Option("\tSet the random number seed." |
---|
| 389 | +"\n\t(default = 1)" |
---|
| 390 | , "S", 1, "-S <seed>")); |
---|
| 391 | newVector.addElement(new Option("\tVerbose output for monitoring the search.","D",0,"-D")); |
---|
| 392 | |
---|
| 393 | return newVector.elements(); |
---|
| 394 | } |
---|
| 395 | |
---|
| 396 | /** |
---|
| 397 | * Parses a given list of options. |
---|
| 398 | * |
---|
| 399 | <!-- options-start --> |
---|
| 400 | * Valid options are: <p> |
---|
| 401 | * |
---|
| 402 | * -Z <br> |
---|
| 403 | * Specify the number of subsets to generate in the initial population.<p> |
---|
| 404 | * |
---|
| 405 | * -T <start set> <br> |
---|
| 406 | * Specify the treshold used for considering when a subset is significant. <p> |
---|
| 407 | * |
---|
| 408 | * -R <br> |
---|
| 409 | * Specify the kind of combiantion. <p> |
---|
| 410 | * |
---|
| 411 | * -S <br> |
---|
| 412 | * Set the random number seed. |
---|
| 413 | * (default = 1) |
---|
| 414 | * |
---|
| 415 | * -D <br> |
---|
| 416 | * Verbose output for monitoring the search |
---|
| 417 | * (default = false) |
---|
| 418 | * |
---|
| 419 | <!-- options-end --> |
---|
| 420 | * |
---|
| 421 | * @param options the list of options as an array of strings |
---|
| 422 | * @exception Exception if an option is not supported |
---|
| 423 | * |
---|
| 424 | **/ |
---|
| 425 | public void setOptions (String[] options) |
---|
| 426 | throws Exception { |
---|
| 427 | String optionString; |
---|
| 428 | resetOptions(); |
---|
| 429 | |
---|
| 430 | optionString = Utils.getOption('Z', options); |
---|
| 431 | if (optionString.length() != 0) { |
---|
| 432 | setPopulationSize(Integer.parseInt(optionString)); |
---|
| 433 | } |
---|
| 434 | |
---|
| 435 | optionString = Utils.getOption('T', options); |
---|
| 436 | if (optionString.length() != 0) { |
---|
| 437 | setTreshold(Double.parseDouble(optionString)); |
---|
| 438 | } |
---|
| 439 | |
---|
| 440 | optionString = Utils.getOption('R', options); |
---|
| 441 | if (optionString.length() != 0) { |
---|
| 442 | setCombination(new SelectedTag(Integer.parseInt(optionString), |
---|
| 443 | TAGS_SELECTION)); |
---|
| 444 | } else { |
---|
| 445 | setCombination(new SelectedTag(COMBINATION_NOT_REDUCED, TAGS_SELECTION)); |
---|
| 446 | } |
---|
| 447 | |
---|
| 448 | optionString = Utils.getOption('S', options); |
---|
| 449 | if (optionString.length() != 0) { |
---|
| 450 | setSeed(Integer.parseInt(optionString)); |
---|
| 451 | } |
---|
| 452 | |
---|
| 453 | setDebug(Utils.getFlag('D', options)); |
---|
| 454 | } |
---|
| 455 | |
---|
| 456 | /** |
---|
| 457 | * Gets the current settings of ScatterSearchV1. |
---|
| 458 | * |
---|
| 459 | * @return an array of strings suitable for passing to setOptions() |
---|
| 460 | */ |
---|
| 461 | public String[] getOptions () { |
---|
| 462 | String[] options = new String[9]; |
---|
| 463 | int current = 0; |
---|
| 464 | |
---|
| 465 | options[current++] = "-T"; |
---|
| 466 | options[current++] = "" + getTreshold (); |
---|
| 467 | |
---|
| 468 | options[current++] = "-Z"; |
---|
| 469 | options[current++] = ""+getPopulationSize (); |
---|
| 470 | |
---|
| 471 | options[current++] = "-R"; |
---|
| 472 | options[current++] = ""+String.valueOf (getCombination ().getSelectedTag ().getID ()); |
---|
| 473 | |
---|
| 474 | options[current++] = "-S"; |
---|
| 475 | options[current++] = "" + getSeed(); |
---|
| 476 | |
---|
| 477 | if (getDebug()) |
---|
| 478 | options[current++] = "-D"; |
---|
| 479 | |
---|
| 480 | while (current < options.length) |
---|
| 481 | options[current++] = ""; |
---|
| 482 | |
---|
| 483 | return options; |
---|
| 484 | } |
---|
| 485 | |
---|
| 486 | /** |
---|
| 487 | * returns a description of the search. |
---|
| 488 | * @return a description of the search as a String. |
---|
| 489 | */ |
---|
| 490 | public String toString() { |
---|
| 491 | StringBuffer FString = new StringBuffer(); |
---|
| 492 | FString.append("\tScatter Search " |
---|
| 493 | + "\n\tInit Population: "+m_calculatedInitialPopSize); |
---|
| 494 | |
---|
| 495 | FString.append("\n\tKind of Combination: " |
---|
| 496 | +getCombination ().getSelectedTag ().getReadable ()); |
---|
| 497 | |
---|
| 498 | FString.append("\n\tRandom number seed: "+m_seed); |
---|
| 499 | |
---|
| 500 | FString.append("\n\tDebug: "+m_debug); |
---|
| 501 | |
---|
| 502 | FString.append("\n\tTreshold: " |
---|
| 503 | +Utils.doubleToString(Math.abs(getTreshold ()),8,3)+"\n"); |
---|
| 504 | |
---|
| 505 | FString.append("\tTotal number of subsets evaluated: " |
---|
| 506 | + m_totalEvals + "\n"); |
---|
| 507 | |
---|
| 508 | FString.append("\tMerit of best subset found: " |
---|
| 509 | +Utils.doubleToString(Math.abs(m_bestMerit),8,3)+"\n"); |
---|
| 510 | |
---|
| 511 | /* FString.append("\tTime procesing the search space: " |
---|
| 512 | +(double)m_processinTime/1000+" seconds\n"); */ |
---|
| 513 | |
---|
| 514 | if(m_debug) |
---|
| 515 | return FString.toString()+"\n\n"+m_InformationReports.toString (); |
---|
| 516 | |
---|
| 517 | return FString.toString(); |
---|
| 518 | } |
---|
| 519 | |
---|
| 520 | |
---|
| 521 | /** |
---|
| 522 | * Searches the attribute subset space using Scatter Search. |
---|
| 523 | * |
---|
| 524 | * @param ASEval the attribute evaluator to guide the search |
---|
| 525 | * @param data the training instances. |
---|
| 526 | * @return an array of selected attribute indexes |
---|
| 527 | * @exception Exception if the search can't be completed |
---|
| 528 | */ |
---|
| 529 | public int[] search(ASEvaluation ASEval, Instances data) |
---|
| 530 | throws Exception{ |
---|
| 531 | |
---|
| 532 | m_totalEvals = 0; |
---|
| 533 | m_popSize = m_initialPopSize; |
---|
| 534 | m_calculatedInitialPopSize = m_initialPopSize; |
---|
| 535 | m_treshold = m_initialThreshold; |
---|
| 536 | m_processinTime =System.currentTimeMillis (); |
---|
| 537 | m_InformationReports = new StringBuffer(); |
---|
| 538 | |
---|
| 539 | m_numAttribs =data.numAttributes (); |
---|
| 540 | m_classIndex =data.classIndex (); |
---|
| 541 | |
---|
| 542 | if(m_popSize<=0) { |
---|
| 543 | m_popSize =m_numAttribs/2; |
---|
| 544 | m_calculatedInitialPopSize = m_popSize; |
---|
| 545 | } |
---|
| 546 | |
---|
| 547 | ASEvaluator =(SubsetEvaluator)ASEval; |
---|
| 548 | |
---|
| 549 | if(!(m_treshold >= 0)){ |
---|
| 550 | m_treshold =calculateTreshhold(); |
---|
| 551 | m_totalEvals++; |
---|
| 552 | } |
---|
| 553 | |
---|
| 554 | m_random = new Random(m_seed); |
---|
| 555 | |
---|
| 556 | m_attributeRanking =RankEachAttribute(); |
---|
| 557 | |
---|
| 558 | CreatePopulation(m_popSize); |
---|
| 559 | |
---|
| 560 | int bestSolutions =m_popSize/4; |
---|
| 561 | int divSolutions =m_popSize/4; |
---|
| 562 | |
---|
| 563 | if(m_popSize < 4){ |
---|
| 564 | |
---|
| 565 | bestSolutions = m_popSize/2; |
---|
| 566 | divSolutions = m_popSize/2; |
---|
| 567 | |
---|
| 568 | if(m_popSize == 1) return attributeList(((Subset)m_population.get (0)).subset); |
---|
| 569 | } |
---|
| 570 | |
---|
| 571 | |
---|
| 572 | m_ReferenceSet =new ArrayList<Subset>(); |
---|
| 573 | |
---|
| 574 | for (int i = 0; i<m_population.size (); i++) { |
---|
| 575 | m_ReferenceSet.add (m_population.get (i)) ; |
---|
| 576 | } |
---|
| 577 | |
---|
| 578 | |
---|
| 579 | GenerateReferenceSet(m_ReferenceSet, bestSolutions, divSolutions); |
---|
| 580 | |
---|
| 581 | |
---|
| 582 | m_InformationReports.append ("Population: "+m_population.size ()+"\n"); |
---|
| 583 | m_InformationReports.append ("merit \tsubset\n"); |
---|
| 584 | |
---|
| 585 | for (int i = 0; i < m_population.size (); i++) |
---|
| 586 | m_InformationReports.append (printSubset (m_population.get (i))); |
---|
| 587 | |
---|
| 588 | |
---|
| 589 | m_ReferenceSet =m_ReferenceSet.subList (0,bestSolutions+divSolutions); |
---|
| 590 | |
---|
| 591 | |
---|
| 592 | /*TEST*/ |
---|
| 593 | m_InformationReports.append ("\nReferenceSet:"); |
---|
| 594 | m_InformationReports.append ("\n----------------Most Significants Solutions--------------\n"); |
---|
| 595 | for (int i = 0; i<m_ReferenceSet.size (); i++) { |
---|
| 596 | if(i ==bestSolutions) m_InformationReports.append ("----------------Most Diverses Solutions--------------\n"); |
---|
| 597 | m_InformationReports.append(printSubset (m_ReferenceSet.get (i))); |
---|
| 598 | } |
---|
| 599 | |
---|
| 600 | |
---|
| 601 | Subset bestTemp =new Subset(new BitSet(m_numAttribs),0); |
---|
| 602 | |
---|
| 603 | while (!(bestTemp.isEqual (m_ReferenceSet.get (0))) /*|| (m_treshold > bestTemp.merit)*/) { |
---|
| 604 | //while(){ |
---|
| 605 | CombineParents(); |
---|
| 606 | ImproveSolutions(); |
---|
| 607 | // } |
---|
| 608 | bestTemp =m_ReferenceSet.get (0); |
---|
| 609 | |
---|
| 610 | int numBest =m_ReferenceSet.size ()/2; |
---|
| 611 | int numDiverses =m_ReferenceSet.size ()/2; |
---|
| 612 | |
---|
| 613 | UpdateReferenceSet(numBest, numDiverses); |
---|
| 614 | m_ReferenceSet = m_ReferenceSet.subList (0,numBest+numDiverses); |
---|
| 615 | |
---|
| 616 | } |
---|
| 617 | |
---|
| 618 | m_InformationReports.append("\nLast Reference Set Updated:\n"); |
---|
| 619 | m_InformationReports.append ("merit \tsubset\n"); |
---|
| 620 | |
---|
| 621 | for (int i = 0; i <m_ReferenceSet.size (); i++) |
---|
| 622 | m_InformationReports.append (printSubset (m_ReferenceSet.get (i))); |
---|
| 623 | |
---|
| 624 | |
---|
| 625 | m_bestMerit =bestTemp.merit; |
---|
| 626 | |
---|
| 627 | m_processinTime =System.currentTimeMillis () -m_processinTime; |
---|
| 628 | |
---|
| 629 | return attributeList (bestTemp.subset); |
---|
| 630 | } |
---|
| 631 | |
---|
| 632 | /** |
---|
| 633 | * Generate the a ReferenceSet containing the n best solutions and the m most diverse solutions of |
---|
| 634 | * the initial Population. |
---|
| 635 | * |
---|
| 636 | * @param ReferenceSet the ReferenceSet for storing these solutions |
---|
| 637 | * @param bestSolutions the number of the most pure solutions. |
---|
| 638 | * @param divSolutions the number of the most diverses solutions acording to the bestSolutions. |
---|
| 639 | */ |
---|
| 640 | public void GenerateReferenceSet(List<Subset> ReferenceSet, int bestSolutions, int divSolutions){ |
---|
| 641 | |
---|
| 642 | //Sorting the Initial ReferenceSet |
---|
| 643 | ReferenceSet =bubbleSubsetSort (ReferenceSet); |
---|
| 644 | |
---|
| 645 | // storing all the attributes that are now in the ReferenceSet (just till bestSolutions) |
---|
| 646 | BitSet allBits_RefSet =getAllBits (ReferenceSet.subList (0,bestSolutions)); |
---|
| 647 | |
---|
| 648 | // for stopping when ReferenceSet.size () ==bestSolutions+divSolutions |
---|
| 649 | int refSetlength =bestSolutions; |
---|
| 650 | int total =bestSolutions+divSolutions; |
---|
| 651 | |
---|
| 652 | while (refSetlength <total) { |
---|
| 653 | |
---|
| 654 | List<Integer> aux =new ArrayList<Integer>(); |
---|
| 655 | |
---|
| 656 | for (int i =refSetlength; i <ReferenceSet.size (); i ++) { |
---|
| 657 | aux.add (SimetricDiference (((Subset)ReferenceSet.get (i)).clone (),allBits_RefSet)); |
---|
| 658 | } |
---|
| 659 | |
---|
| 660 | |
---|
| 661 | int mostDiv =getIndexofBiggest(aux); |
---|
| 662 | ReferenceSet.set(refSetlength, ReferenceSet.get (refSetlength+mostDiv)); |
---|
| 663 | //ReferenceSet.remove (refSetlength +mostDiv); |
---|
| 664 | |
---|
| 665 | refSetlength++; |
---|
| 666 | |
---|
| 667 | allBits_RefSet =getAllBits (ReferenceSet.subList (0,refSetlength)); |
---|
| 668 | } |
---|
| 669 | |
---|
| 670 | ReferenceSet =filterSubset (ReferenceSet,refSetlength); |
---|
| 671 | } |
---|
| 672 | |
---|
| 673 | /** |
---|
| 674 | * Update the ReferenceSet putting the new obtained Solutions there |
---|
| 675 | * |
---|
| 676 | * @param numBestSolutions the number of the most pure solutions. |
---|
| 677 | * @param numDivsSolutions the number of the most diverses solutions acording to the bestSolutions. |
---|
| 678 | */ |
---|
| 679 | public void UpdateReferenceSet(int numBestSolutions, int numDivsSolutions){ |
---|
| 680 | |
---|
| 681 | for (int i = 0; i <m_parentsCombination.size (); i++) m_ReferenceSet.add (i, m_parentsCombination.get (i)); |
---|
| 682 | |
---|
| 683 | GenerateReferenceSet (m_ReferenceSet,numBestSolutions,numDivsSolutions); |
---|
| 684 | } |
---|
| 685 | |
---|
| 686 | /** |
---|
| 687 | * Improve the solutions previously combined by adding the attributes that improve that solution |
---|
| 688 | * @exception Exception if there is some trouble evaluating the candidate solutions |
---|
| 689 | */ |
---|
| 690 | public void ImproveSolutions() |
---|
| 691 | throws Exception{ |
---|
| 692 | |
---|
| 693 | for (int i = 0; i<m_parentsCombination.size (); i++) { |
---|
| 694 | |
---|
| 695 | BitSet aux1 =(BitSet)((Subset)m_parentsCombination.get (i)).subset.clone (); |
---|
| 696 | List<Subset> ranking =new ArrayList<Subset>(); |
---|
| 697 | |
---|
| 698 | /* |
---|
| 699 | for(int j=aux1.nextClearBit (0); j<=m_numAttribs; j=aux1.nextClearBit(j+1)){ |
---|
| 700 | if(j ==m_classIndex)continue; |
---|
| 701 | |
---|
| 702 | BitSet aux2 =new BitSet(m_numAttribs); |
---|
| 703 | aux2.set (j); |
---|
| 704 | |
---|
| 705 | double merit =ASEvaluator.evaluateSubset (aux2); |
---|
| 706 | m_totalEvals++; |
---|
| 707 | |
---|
| 708 | ranking.add (new Subset((BitSet)aux2.clone (), merit)); |
---|
| 709 | } |
---|
| 710 | |
---|
| 711 | ranking =bubbleSubsetSort (ranking); |
---|
| 712 | */ |
---|
| 713 | |
---|
| 714 | for (int k =0; k <m_attributeRanking.size (); k ++) { |
---|
| 715 | Subset s1 =((Subset)m_attributeRanking.get (k)).clone (); |
---|
| 716 | BitSet b1 =(BitSet)s1.subset.clone (); |
---|
| 717 | |
---|
| 718 | Subset s2 =((Subset)m_parentsCombination.get (i)).clone (); |
---|
| 719 | BitSet b2 =(BitSet)s2.subset.clone (); |
---|
| 720 | |
---|
| 721 | if(b2.get (b1.nextSetBit (0))) continue; |
---|
| 722 | |
---|
| 723 | b2.or (b1); |
---|
| 724 | double newMerit =ASEvaluator.evaluateSubset (b2); |
---|
| 725 | m_totalEvals++; |
---|
| 726 | |
---|
| 727 | if(newMerit <= s2.merit)break; |
---|
| 728 | |
---|
| 729 | m_parentsCombination.set (i,new Subset(b2,newMerit)); |
---|
| 730 | } |
---|
| 731 | |
---|
| 732 | filterSubset (m_parentsCombination,m_ReferenceSet.size()); |
---|
| 733 | } |
---|
| 734 | } |
---|
| 735 | |
---|
| 736 | /** |
---|
| 737 | * Combine all the posible pair solutions existing in the Population |
---|
| 738 | * |
---|
| 739 | * @exception Exception if there is some trouble evaluating the new childs |
---|
| 740 | */ |
---|
| 741 | public void CombineParents() |
---|
| 742 | throws Exception{ |
---|
| 743 | |
---|
| 744 | m_parentsCombination =new ArrayList<Subset>(); |
---|
| 745 | |
---|
| 746 | // this two 'for' are for selecting parents in the refSet |
---|
| 747 | for (int i= 0; i <m_ReferenceSet.size ()-1; i ++) { |
---|
| 748 | for (int j= i+1; j <m_ReferenceSet.size (); j ++) { |
---|
| 749 | |
---|
| 750 | // Selecting parents |
---|
| 751 | Subset parent1 =m_ReferenceSet.get (i); |
---|
| 752 | Subset parent2 =m_ReferenceSet.get (j); |
---|
| 753 | |
---|
| 754 | // Initializing childs Intersecting parents |
---|
| 755 | Subset child1 = intersectSubsets (parent1, parent2); |
---|
| 756 | Subset child2 =child1.clone (); |
---|
| 757 | |
---|
| 758 | // Initializing childs Intersecting parents |
---|
| 759 | Subset simDif =simetricDif (parent1, parent2, getCombination ().getSelectedTag ().getID ()); |
---|
| 760 | |
---|
| 761 | BitSet aux =(BitSet)simDif.subset.clone (); |
---|
| 762 | |
---|
| 763 | boolean improvement =true; |
---|
| 764 | |
---|
| 765 | while (improvement) { |
---|
| 766 | |
---|
| 767 | Subset best1 =getBestgen (child1,aux); |
---|
| 768 | Subset best2 =getBestgen (child2,aux); |
---|
| 769 | |
---|
| 770 | if(best1 !=null || best2!=null){ |
---|
| 771 | |
---|
| 772 | if(best2 ==null){ |
---|
| 773 | child1 =best1.clone (); |
---|
| 774 | continue; |
---|
| 775 | } |
---|
| 776 | if(best1 ==null){ |
---|
| 777 | child2 =best2.clone (); |
---|
| 778 | continue; |
---|
| 779 | } |
---|
| 780 | if(best1 !=null && best2 !=null){ |
---|
| 781 | double merit1 =best1.merit; |
---|
| 782 | double merit2 =best2.merit; |
---|
| 783 | |
---|
| 784 | if(merit1 >merit2){ |
---|
| 785 | child1 =best1.clone (); |
---|
| 786 | continue; |
---|
| 787 | } |
---|
| 788 | if(merit1 <merit2){ |
---|
| 789 | child2 =best2.clone (); |
---|
| 790 | continue; |
---|
| 791 | } |
---|
| 792 | if(merit1 ==merit2){ |
---|
| 793 | if(best1.subset.cardinality () > best2.subset.cardinality ()){ |
---|
| 794 | child2 =best2.clone (); |
---|
| 795 | continue; |
---|
| 796 | } |
---|
| 797 | if(best1.subset.cardinality () < best2.subset.cardinality ()){ |
---|
| 798 | child1 =best1.clone (); |
---|
| 799 | continue; |
---|
| 800 | } |
---|
| 801 | if(best1.subset.cardinality () == best2.subset.cardinality ()){ |
---|
| 802 | double random = m_random.nextDouble (); |
---|
| 803 | if(random < 0.5)child1 =best1.clone (); |
---|
| 804 | else child2 =best2.clone (); |
---|
| 805 | continue; |
---|
| 806 | } |
---|
| 807 | } |
---|
| 808 | } |
---|
| 809 | |
---|
| 810 | }else{ |
---|
| 811 | m_parentsCombination.add (child1); |
---|
| 812 | m_parentsCombination.add (child2); |
---|
| 813 | improvement =false; |
---|
| 814 | } |
---|
| 815 | } |
---|
| 816 | } |
---|
| 817 | } |
---|
| 818 | m_parentsCombination = filterSubset (m_parentsCombination,m_ReferenceSet.size()); |
---|
| 819 | |
---|
| 820 | GenerateReferenceSet (m_parentsCombination,m_ReferenceSet.size ()/2, m_ReferenceSet.size ()/2); |
---|
| 821 | m_parentsCombination = m_parentsCombination.subList (0, m_ReferenceSet.size ()); |
---|
| 822 | |
---|
| 823 | } |
---|
| 824 | /** |
---|
| 825 | * Create the initial Population |
---|
| 826 | * |
---|
| 827 | * @param popSize the size of the initial population |
---|
| 828 | * @exception Exception if there is a trouble evaluating any solution |
---|
| 829 | */ |
---|
| 830 | public void CreatePopulation(int popSize) |
---|
| 831 | throws Exception{ |
---|
| 832 | |
---|
| 833 | InitPopulation(popSize); |
---|
| 834 | |
---|
| 835 | /** Delimit the best attributes from the worst*/ |
---|
| 836 | int segmentation =m_numAttribs/2; |
---|
| 837 | |
---|
| 838 | /*TEST*/ |
---|
| 839 | /* System.out.println ("AttributeRanking"); |
---|
| 840 | for (int i = 0; i <attributeRanking.size (); i++){ |
---|
| 841 | if(i ==segmentation)System.out.println ("-------------------------SEGMENTATION------------------------"); |
---|
| 842 | printSubset (attributeRanking.get (i)); |
---|
| 843 | } |
---|
| 844 | */ |
---|
| 845 | for (int i = 0; i<m_popSize; i++) { |
---|
| 846 | |
---|
| 847 | List<Subset> attributeRankingCopy = new ArrayList<Subset>(); |
---|
| 848 | for (int j = 0; j<m_attributeRanking.size (); j++) attributeRankingCopy.add (m_attributeRanking.get (j)); |
---|
| 849 | |
---|
| 850 | |
---|
| 851 | double last_evaluation =-999; |
---|
| 852 | double current_evaluation =0; |
---|
| 853 | |
---|
| 854 | boolean doneAnew =true; |
---|
| 855 | |
---|
| 856 | while (true) { |
---|
| 857 | |
---|
| 858 | // generate a random number in the interval[0..segmentation] |
---|
| 859 | int random_number = m_random.nextInt (segmentation+1) /*generateRandomNumber (segmentation)*/; |
---|
| 860 | |
---|
| 861 | if(doneAnew && i <=segmentation)random_number =i; |
---|
| 862 | doneAnew =false; |
---|
| 863 | |
---|
| 864 | Subset s1 =((Subset)attributeRankingCopy.get (random_number)).clone (); |
---|
| 865 | Subset s2 =((Subset)m_population.get (i)).clone (); |
---|
| 866 | |
---|
| 867 | |
---|
| 868 | // trying to add a new gen in the chromosome i of the population |
---|
| 869 | Subset joiners =joinSubsets (s1, s2 ); |
---|
| 870 | |
---|
| 871 | current_evaluation =joiners.merit; |
---|
| 872 | |
---|
| 873 | if(current_evaluation > last_evaluation){ |
---|
| 874 | m_population.set (i,joiners); |
---|
| 875 | last_evaluation =current_evaluation; |
---|
| 876 | |
---|
| 877 | try { |
---|
| 878 | attributeRankingCopy.set (random_number, attributeRankingCopy.get (segmentation+1)); |
---|
| 879 | attributeRankingCopy.remove (segmentation+1); |
---|
| 880 | }catch (IndexOutOfBoundsException ex) { |
---|
| 881 | attributeRankingCopy.set (random_number,new Subset(new BitSet(m_numAttribs),0)); |
---|
| 882 | continue; |
---|
| 883 | } |
---|
| 884 | } |
---|
| 885 | else{ |
---|
| 886 | // there's not more improvement |
---|
| 887 | break; |
---|
| 888 | } |
---|
| 889 | |
---|
| 890 | |
---|
| 891 | } |
---|
| 892 | } |
---|
| 893 | |
---|
| 894 | //m_population =bubbleSubsetSort (m_population); |
---|
| 895 | } |
---|
| 896 | |
---|
| 897 | |
---|
| 898 | /** |
---|
| 899 | * Rank all the attributes individually acording to their merits |
---|
| 900 | * |
---|
| 901 | * @return an ordered List of Subsets with just one attribute |
---|
| 902 | * @exception Exception if the evaluation can not be completed |
---|
| 903 | */ |
---|
| 904 | public List<Subset> RankEachAttribute() |
---|
| 905 | throws Exception{ |
---|
| 906 | |
---|
| 907 | List<Subset> result =new ArrayList<Subset>(); |
---|
| 908 | |
---|
| 909 | for (int i = 0; i<m_numAttribs; i++) { |
---|
| 910 | if(i==m_classIndex)continue; |
---|
| 911 | |
---|
| 912 | BitSet an_Attribute =new BitSet(m_numAttribs); |
---|
| 913 | an_Attribute.set (i); |
---|
| 914 | |
---|
| 915 | double merit =ASEvaluator.evaluateSubset (an_Attribute); |
---|
| 916 | m_totalEvals++; |
---|
| 917 | |
---|
| 918 | result.add (new Subset(an_Attribute, merit)); |
---|
| 919 | } |
---|
| 920 | |
---|
| 921 | return bubbleSubsetSort(result); |
---|
| 922 | } |
---|
| 923 | |
---|
| 924 | |
---|
| 925 | //.......... |
---|
| 926 | |
---|
| 927 | |
---|
| 928 | /** |
---|
| 929 | * Evaluate each gen of a BitSet inserted in a Subset and get the most significant for that Subset |
---|
| 930 | * |
---|
| 931 | * @return a new Subset with the union of subset and the best gen of gens. |
---|
| 932 | * in case that there's not improvement with each gen return null |
---|
| 933 | * @exception Exception if the evaluation of can not be completed |
---|
| 934 | */ |
---|
| 935 | public Subset getBestgen(Subset subset, BitSet gens) |
---|
| 936 | throws Exception{ |
---|
| 937 | Subset result =null; |
---|
| 938 | |
---|
| 939 | double merit1 =subset.merit; |
---|
| 940 | |
---|
| 941 | for(int i =gens.nextSetBit(0); i >=0; i =gens.nextSetBit(i+1)){ |
---|
| 942 | BitSet aux =(BitSet)subset.subset.clone (); |
---|
| 943 | |
---|
| 944 | if(aux.get (i))continue; |
---|
| 945 | aux.set (i); |
---|
| 946 | |
---|
| 947 | double merit2 =ASEvaluator.evaluateSubset (aux); |
---|
| 948 | m_totalEvals++; |
---|
| 949 | |
---|
| 950 | if(merit2 >merit1){ |
---|
| 951 | merit1 =merit2; |
---|
| 952 | result =new Subset(aux,merit1); |
---|
| 953 | } |
---|
| 954 | } |
---|
| 955 | |
---|
| 956 | return result; |
---|
| 957 | } |
---|
| 958 | |
---|
| 959 | /** |
---|
| 960 | * Sort a List of subsets according to their merits |
---|
| 961 | * |
---|
| 962 | * @param subsetList the subsetList to be ordered |
---|
| 963 | * @return a List with ordered subsets |
---|
| 964 | */ |
---|
| 965 | public List<Subset> bubbleSubsetSort(List<Subset> subsetList){ |
---|
| 966 | List<Subset> result =new ArrayList<Subset>(); |
---|
| 967 | |
---|
| 968 | for (int i = 0; i<subsetList.size ()-1; i++) { |
---|
| 969 | Subset subset1 =subsetList.get (i); |
---|
| 970 | double merit1 =subset1.merit; |
---|
| 971 | |
---|
| 972 | for (int j = i+1; j<subsetList.size (); j++) { |
---|
| 973 | Subset subset2 =subsetList.get (j); |
---|
| 974 | double merit2 =subset2.merit; |
---|
| 975 | |
---|
| 976 | if(merit2 > merit1){ |
---|
| 977 | Subset temp =subset1; |
---|
| 978 | |
---|
| 979 | subsetList.set (i,subset2); |
---|
| 980 | subsetList.set (j,temp); |
---|
| 981 | |
---|
| 982 | subset1 =subset2; |
---|
| 983 | merit1 =subset1.merit; |
---|
| 984 | } |
---|
| 985 | } |
---|
| 986 | } |
---|
| 987 | return subsetList; |
---|
| 988 | } |
---|
| 989 | |
---|
| 990 | |
---|
| 991 | /** |
---|
| 992 | * get the index in a List where this have the biggest number |
---|
| 993 | * |
---|
| 994 | * @param simDif the Lists of numbers for getting from them the index of the bigger |
---|
| 995 | * @return an index that represents where the bigest number is. |
---|
| 996 | */ |
---|
| 997 | public int getIndexofBiggest(List<Integer> simDif){ |
---|
| 998 | int aux =-99999; |
---|
| 999 | int result1 =-1; |
---|
| 1000 | List<Integer> equalSimDif =new ArrayList<Integer>(); |
---|
| 1001 | |
---|
| 1002 | if(simDif.size ()==0) return -1; |
---|
| 1003 | |
---|
| 1004 | for (int i = 0; i<simDif.size (); i++) { |
---|
| 1005 | if(simDif.get (i) >aux){ |
---|
| 1006 | aux =simDif.get (i); |
---|
| 1007 | result1 =i; |
---|
| 1008 | } |
---|
| 1009 | } |
---|
| 1010 | |
---|
| 1011 | for (int i =0; i <simDif.size (); i++) { |
---|
| 1012 | if(simDif.get (i) ==aux){ |
---|
| 1013 | equalSimDif.add (i); |
---|
| 1014 | } |
---|
| 1015 | } |
---|
| 1016 | |
---|
| 1017 | int finalResult =equalSimDif.get (m_random.nextInt (equalSimDif.size ()) /*generateRandomNumber (equalSimDif.size ()-1)*/); |
---|
| 1018 | |
---|
| 1019 | return finalResult; |
---|
| 1020 | } |
---|
| 1021 | |
---|
| 1022 | /** |
---|
| 1023 | * Save in Bitset all the gens that are in many others subsets. |
---|
| 1024 | * |
---|
| 1025 | * @param subsets the Lists of subsets for getting from them all their gens |
---|
| 1026 | * @return a Bitset with all the gens contained in many others subsets. |
---|
| 1027 | */ |
---|
| 1028 | public BitSet getAllBits(List<Subset> subsets){ |
---|
| 1029 | BitSet result =new BitSet(m_numAttribs); |
---|
| 1030 | |
---|
| 1031 | for (int i =0; i <subsets.size (); i ++) { |
---|
| 1032 | BitSet aux =((Subset)subsets.get (i)).clone ().subset; |
---|
| 1033 | |
---|
| 1034 | for(int j=aux.nextSetBit(0); j>=0; j=aux.nextSetBit(j+1)) { |
---|
| 1035 | result.set (j); |
---|
| 1036 | } |
---|
| 1037 | } |
---|
| 1038 | |
---|
| 1039 | return result; |
---|
| 1040 | } |
---|
| 1041 | |
---|
| 1042 | /** |
---|
| 1043 | * Creating space for introducing the population |
---|
| 1044 | * |
---|
| 1045 | * @param popSize the number of subset in the initial population |
---|
| 1046 | */ |
---|
| 1047 | public void InitPopulation(int popSize){ |
---|
| 1048 | m_population =new ArrayList<Subset>(); |
---|
| 1049 | for (int i = 0; i<popSize; i++)m_population.add (new Subset(new BitSet(m_numAttribs),0)); |
---|
| 1050 | } |
---|
| 1051 | |
---|
| 1052 | /** |
---|
| 1053 | * Join two subsets |
---|
| 1054 | * |
---|
| 1055 | * @param subset1 one of the subsets |
---|
| 1056 | * @param subset2 the other subset |
---|
| 1057 | * @return a new Subset that is te result of the Join |
---|
| 1058 | * @exception Exception if the evaluation of the subsets can not be completed |
---|
| 1059 | */ |
---|
| 1060 | public Subset joinSubsets(Subset subset1, Subset subset2) |
---|
| 1061 | throws Exception{ |
---|
| 1062 | BitSet b1 =(BitSet)subset1.subset.clone (); |
---|
| 1063 | BitSet b2 =(BitSet)subset2.subset.clone (); |
---|
| 1064 | |
---|
| 1065 | b1.or (b2); |
---|
| 1066 | |
---|
| 1067 | double newMerit =ASEvaluator.evaluateSubset (b1); |
---|
| 1068 | m_totalEvals++; |
---|
| 1069 | |
---|
| 1070 | return new Subset((BitSet)b1.clone (), newMerit); |
---|
| 1071 | } |
---|
| 1072 | |
---|
| 1073 | /** |
---|
| 1074 | * Intersects two subsets |
---|
| 1075 | * |
---|
| 1076 | * @param subset1 one of the subsets |
---|
| 1077 | * @param subset2 the other subset |
---|
| 1078 | * @return a new Subset that is te result of the intersection |
---|
| 1079 | * @exception Exception if the evaluation of the subsets can not be completed |
---|
| 1080 | */ |
---|
| 1081 | public Subset intersectSubsets(Subset subset1, Subset subset2) |
---|
| 1082 | throws Exception{ |
---|
| 1083 | BitSet b1 =(BitSet)subset1.subset.clone (); |
---|
| 1084 | BitSet b2 =(BitSet)subset2.subset.clone (); |
---|
| 1085 | |
---|
| 1086 | b1.and (b2); |
---|
| 1087 | |
---|
| 1088 | double newMerit =ASEvaluator.evaluateSubset (b1); |
---|
| 1089 | m_totalEvals++; |
---|
| 1090 | |
---|
| 1091 | return new Subset((BitSet)b1.clone (), newMerit); |
---|
| 1092 | } |
---|
| 1093 | |
---|
| 1094 | public Subset simetricDif(Subset subset1, Subset subset2, int mode) |
---|
| 1095 | throws Exception{ |
---|
| 1096 | BitSet b1 =(BitSet)subset1.subset.clone (); |
---|
| 1097 | BitSet b2 =(BitSet)subset2.subset.clone (); |
---|
| 1098 | |
---|
| 1099 | b1.xor (b2); |
---|
| 1100 | |
---|
| 1101 | double newMerit =ASEvaluator.evaluateSubset (b1); |
---|
| 1102 | m_totalEvals++; |
---|
| 1103 | |
---|
| 1104 | Subset result =new Subset((BitSet)b1.clone (), newMerit); |
---|
| 1105 | |
---|
| 1106 | if(mode == COMBINATION_REDUCED){ |
---|
| 1107 | |
---|
| 1108 | double avgAcurracy =0; |
---|
| 1109 | int totalSolutions =0; |
---|
| 1110 | List<Subset> weightVector =new ArrayList<Subset>(); |
---|
| 1111 | |
---|
| 1112 | BitSet res =result.subset; |
---|
| 1113 | for(int i=res.nextSetBit(0); i>=0; i=res.nextSetBit(i+1)){ |
---|
| 1114 | |
---|
| 1115 | double merits =0; |
---|
| 1116 | int numSolutions =0; |
---|
| 1117 | Subset solution =null; |
---|
| 1118 | |
---|
| 1119 | for (int j = 0; j <m_ReferenceSet.size (); j ++) { |
---|
| 1120 | solution =(Subset)m_ReferenceSet.get (j); |
---|
| 1121 | if(solution.subset.get (i)){ |
---|
| 1122 | merits +=solution.merit; |
---|
| 1123 | numSolutions ++; |
---|
| 1124 | } |
---|
| 1125 | } |
---|
| 1126 | BitSet b =new BitSet(m_numAttribs); |
---|
| 1127 | b.set (i); |
---|
| 1128 | Subset s =new Subset(b, merits/(double)numSolutions); |
---|
| 1129 | weightVector.add (s); |
---|
| 1130 | |
---|
| 1131 | avgAcurracy +=merits; |
---|
| 1132 | totalSolutions ++; |
---|
| 1133 | |
---|
| 1134 | } |
---|
| 1135 | avgAcurracy =avgAcurracy/(double)totalSolutions; |
---|
| 1136 | |
---|
| 1137 | BitSet newResult =new BitSet(m_numAttribs); |
---|
| 1138 | for (int i = 0; i<weightVector.size (); i++) { |
---|
| 1139 | Subset aux =weightVector.get (i); |
---|
| 1140 | if(aux.merit >=avgAcurracy){ |
---|
| 1141 | newResult.or (aux.subset); |
---|
| 1142 | } |
---|
| 1143 | } |
---|
| 1144 | double merit =ASEvaluator.evaluateSubset (newResult); |
---|
| 1145 | result =new Subset(newResult, merit); |
---|
| 1146 | |
---|
| 1147 | } |
---|
| 1148 | |
---|
| 1149 | return result; |
---|
| 1150 | } |
---|
| 1151 | |
---|
| 1152 | public int generateRandomNumber(int limit){ |
---|
| 1153 | |
---|
| 1154 | return (int)Math.round (Math.random ()*(limit+0.4)); |
---|
| 1155 | } |
---|
| 1156 | |
---|
| 1157 | |
---|
| 1158 | |
---|
| 1159 | /** |
---|
| 1160 | * Calculate the treshold of a dataSet given an evaluator |
---|
| 1161 | * |
---|
| 1162 | * @return the treshhold of the dataSet |
---|
| 1163 | * @exception Exception if the calculation can not be completed |
---|
| 1164 | */ |
---|
| 1165 | public double calculateTreshhold() |
---|
| 1166 | throws Exception{ |
---|
| 1167 | BitSet fullSet =new BitSet(m_numAttribs); |
---|
| 1168 | |
---|
| 1169 | for (int i= 0; i< m_numAttribs; i++) { |
---|
| 1170 | if(i ==m_classIndex)continue; |
---|
| 1171 | fullSet.set (i); |
---|
| 1172 | } |
---|
| 1173 | |
---|
| 1174 | return ASEvaluator.evaluateSubset (fullSet); |
---|
| 1175 | } |
---|
| 1176 | |
---|
| 1177 | /** |
---|
| 1178 | * Calculate the Simetric Diference of two subsets |
---|
| 1179 | * |
---|
| 1180 | * @return the Simetric Diference |
---|
| 1181 | * @exception Exception if the calculation can not be completed |
---|
| 1182 | */ |
---|
| 1183 | public int SimetricDiference(Subset subset, BitSet bitset){ |
---|
| 1184 | BitSet aux =subset.clone ().subset; |
---|
| 1185 | aux.xor (bitset); |
---|
| 1186 | |
---|
| 1187 | return aux.cardinality (); |
---|
| 1188 | } |
---|
| 1189 | |
---|
| 1190 | /** |
---|
| 1191 | * Filter a given Lis of Subsets removing the equals subsets |
---|
| 1192 | * @param subsetList to filter |
---|
| 1193 | * @param preferredSize the preferred size of the new List (if it is -1, then the filter is make it |
---|
| 1194 | * for all subsets, else then the filter method stops when the given preferred |
---|
| 1195 | * size is reached or all the subset have been filtered). |
---|
| 1196 | * @return a new List filtered |
---|
| 1197 | * @exception Exception if the calculation can not be completed |
---|
| 1198 | */ |
---|
| 1199 | public List<Subset> filterSubset(List<Subset> subsetList, int preferredSize){ |
---|
| 1200 | if(subsetList.size () <=preferredSize && preferredSize !=-1)return subsetList; |
---|
| 1201 | |
---|
| 1202 | for (int i =0; i <subsetList.size ()-1; i ++) { |
---|
| 1203 | for (int j =i+1; j <subsetList.size (); j ++) { |
---|
| 1204 | Subset focus =subsetList.get (i); |
---|
| 1205 | if(focus.isEqual (subsetList.get (j))){ |
---|
| 1206 | subsetList.remove (j); |
---|
| 1207 | j--; |
---|
| 1208 | |
---|
| 1209 | if(subsetList.size () <=preferredSize && preferredSize !=-1)return subsetList; |
---|
| 1210 | } |
---|
| 1211 | } |
---|
| 1212 | } |
---|
| 1213 | return subsetList; |
---|
| 1214 | } |
---|
| 1215 | //.......... |
---|
| 1216 | |
---|
| 1217 | |
---|
| 1218 | // Test Methods |
---|
| 1219 | |
---|
| 1220 | public String printSubset(Subset subset){ |
---|
| 1221 | StringBuffer bufferString = new StringBuffer(); |
---|
| 1222 | |
---|
| 1223 | if(subset == null){ |
---|
| 1224 | //System.out.println ("null"); |
---|
| 1225 | return ""; |
---|
| 1226 | } |
---|
| 1227 | |
---|
| 1228 | BitSet bits =subset.subset; |
---|
| 1229 | double merit =subset.merit; |
---|
| 1230 | List<Integer> indexes =new ArrayList<Integer>(); |
---|
| 1231 | |
---|
| 1232 | for (int i = 0; i<m_numAttribs; i++) { |
---|
| 1233 | if(bits.get (i)){ |
---|
| 1234 | //System.out.print ("1"); |
---|
| 1235 | indexes.add (i+1); |
---|
| 1236 | } |
---|
| 1237 | //else System.out.print ("0"); |
---|
| 1238 | } |
---|
| 1239 | bufferString.append (Utils.doubleToString (merit,8,5)+"\t "+indexes.toString ()+"\n"); |
---|
| 1240 | //System.out.print (" with a merit of: "+merit); |
---|
| 1241 | |
---|
| 1242 | return bufferString.toString (); |
---|
| 1243 | } |
---|
| 1244 | |
---|
| 1245 | //........ |
---|
| 1246 | |
---|
| 1247 | protected void resetOptions () { |
---|
| 1248 | m_popSize = -1; |
---|
| 1249 | m_initialPopSize = -1; |
---|
| 1250 | m_calculatedInitialPopSize = -1; |
---|
| 1251 | m_treshold = -1; |
---|
| 1252 | m_typeOfCombination = COMBINATION_NOT_REDUCED; |
---|
| 1253 | m_seed = 1; |
---|
| 1254 | m_debug = true; |
---|
| 1255 | m_totalEvals = 0; |
---|
| 1256 | m_bestMerit = 0; |
---|
| 1257 | m_processinTime = 0; |
---|
| 1258 | } |
---|
| 1259 | |
---|
| 1260 | /** |
---|
| 1261 | * converts a BitSet into a list of attribute indexes |
---|
| 1262 | * @param group the BitSet to convert |
---|
| 1263 | * @return an array of attribute indexes |
---|
| 1264 | **/ |
---|
| 1265 | public int[] attributeList (BitSet group) { |
---|
| 1266 | int count = 0; |
---|
| 1267 | |
---|
| 1268 | // count how many were selected |
---|
| 1269 | for (int i = 0; i < m_numAttribs; i++) { |
---|
| 1270 | if (group.get(i)) { |
---|
| 1271 | count++; |
---|
| 1272 | } |
---|
| 1273 | } |
---|
| 1274 | |
---|
| 1275 | int[] list = new int[count]; |
---|
| 1276 | count = 0; |
---|
| 1277 | |
---|
| 1278 | for (int i = 0; i < m_numAttribs; i++) { |
---|
| 1279 | if (group.get(i)) { |
---|
| 1280 | list[count++] = i; |
---|
| 1281 | } |
---|
| 1282 | } |
---|
| 1283 | |
---|
| 1284 | return list; |
---|
| 1285 | } |
---|
| 1286 | |
---|
| 1287 | |
---|
| 1288 | // Auxiliar Class for handling Chromosomes and its respective merit |
---|
| 1289 | |
---|
| 1290 | public class Subset implements Serializable { |
---|
| 1291 | |
---|
| 1292 | double merit; |
---|
| 1293 | BitSet subset; |
---|
| 1294 | |
---|
| 1295 | public Subset(BitSet subset, double merit){ |
---|
| 1296 | this.subset =(BitSet)subset.clone (); |
---|
| 1297 | this.merit =merit; |
---|
| 1298 | } |
---|
| 1299 | |
---|
| 1300 | public boolean isEqual(Subset othersubset){ |
---|
| 1301 | if(subset.equals (othersubset.subset))return true; |
---|
| 1302 | return false; |
---|
| 1303 | } |
---|
| 1304 | |
---|
| 1305 | public Subset clone(){ |
---|
| 1306 | return new Subset((BitSet)subset.clone (), merit); |
---|
| 1307 | } |
---|
| 1308 | } |
---|
| 1309 | //.......... |
---|
| 1310 | |
---|
| 1311 | } |
---|
| 1312 | |
---|