[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 | * Rule.java |
---|
| 19 | * Copyright (C) 2003 Peter A. Flach, Nicolas Lachiche |
---|
| 20 | * |
---|
| 21 | * Thanks to Amelie Deltour for porting the original C code to Java |
---|
| 22 | * and integrating it into Weka. |
---|
| 23 | */ |
---|
| 24 | |
---|
| 25 | package weka.associations.tertius; |
---|
| 26 | |
---|
| 27 | import weka.core.Instance; |
---|
| 28 | import weka.core.Instances; |
---|
| 29 | import weka.core.RevisionHandler; |
---|
| 30 | import weka.core.RevisionUtils; |
---|
| 31 | |
---|
| 32 | import java.io.Serializable; |
---|
| 33 | import java.text.DecimalFormat; |
---|
| 34 | import java.util.ArrayList; |
---|
| 35 | import java.util.Comparator; |
---|
| 36 | import java.util.Enumeration; |
---|
| 37 | |
---|
| 38 | /** |
---|
| 39 | * Class representing a rule with a body and a head. |
---|
| 40 | * |
---|
| 41 | * @author <a href="mailto:adeltour@netcourrier.com">Amelie Deltour</a> |
---|
| 42 | * @version $Revision: 1.7 $ |
---|
| 43 | */ |
---|
| 44 | |
---|
| 45 | public class Rule |
---|
| 46 | implements Serializable, Cloneable, RevisionHandler { |
---|
| 47 | |
---|
| 48 | /** for serialization */ |
---|
| 49 | private static final long serialVersionUID = -7763378359090435505L; |
---|
| 50 | |
---|
| 51 | /** The body of the rule. */ |
---|
| 52 | private Body m_body; |
---|
| 53 | |
---|
| 54 | /** The head of the rule. */ |
---|
| 55 | private Head m_head; |
---|
| 56 | |
---|
| 57 | /** Can repeat predicates in the rule ? */ |
---|
| 58 | private boolean m_repeatPredicate; |
---|
| 59 | |
---|
| 60 | /** Maximal number of literals in the rule. */ |
---|
| 61 | private int m_maxLiterals; |
---|
| 62 | |
---|
| 63 | /** Can there be negations in the body ? */ |
---|
| 64 | private boolean m_negBody; |
---|
| 65 | |
---|
| 66 | /** Can there be negations in the head ? */ |
---|
| 67 | private boolean m_negHead; |
---|
| 68 | |
---|
| 69 | /** Is this rule a classification rule ? */ |
---|
| 70 | private boolean m_classRule; |
---|
| 71 | |
---|
| 72 | /** Can there be only one literal in the head ? */ |
---|
| 73 | private boolean m_singleHead; |
---|
| 74 | |
---|
| 75 | /** Number of instances in the data this rule deals with. */ |
---|
| 76 | private int m_numInstances; |
---|
| 77 | |
---|
| 78 | /** Set of counter-instances of this rule. */ |
---|
| 79 | private ArrayList m_counterInstances; |
---|
| 80 | |
---|
| 81 | /** Counter for the counter-instances of this rule. */ |
---|
| 82 | private int m_counter; |
---|
| 83 | |
---|
| 84 | /** Confirmation of this rule. */ |
---|
| 85 | private double m_confirmation; |
---|
| 86 | |
---|
| 87 | /** Optimistic estimate of this rule. */ |
---|
| 88 | private double m_optimistic; |
---|
| 89 | |
---|
| 90 | /** |
---|
| 91 | * Constructor for a rule when the counter-instances are not stored, |
---|
| 92 | * giving all the constraints applied to this rule. |
---|
| 93 | * |
---|
| 94 | * @param repeatPredicate True if predicates can be repeated. |
---|
| 95 | * @param maxLiterals Maximum number of literals. |
---|
| 96 | * @param negBody True if negation is allowed in the body. |
---|
| 97 | * @param negHead True if negation is allowed in the head. |
---|
| 98 | * @param classRule True if the rule is a classification rule. |
---|
| 99 | * @param horn True if the rule is a horn clause. |
---|
| 100 | */ |
---|
| 101 | public Rule(boolean repeatPredicate, int maxLiterals, boolean negBody, |
---|
| 102 | boolean negHead, boolean classRule, boolean horn) { |
---|
| 103 | |
---|
| 104 | m_body = new Body(); |
---|
| 105 | m_head = new Head(); |
---|
| 106 | m_repeatPredicate = repeatPredicate; |
---|
| 107 | m_maxLiterals = maxLiterals; |
---|
| 108 | m_negBody = negBody && !horn; |
---|
| 109 | m_negHead = negHead && !horn; |
---|
| 110 | m_classRule = classRule; |
---|
| 111 | m_singleHead = classRule || horn; |
---|
| 112 | } |
---|
| 113 | |
---|
| 114 | /** |
---|
| 115 | * Constructor for a rule when the counter-instances are stored, |
---|
| 116 | * giving all the constraints applied to this rule. |
---|
| 117 | * The counter-instances are initialized to all the instances in the dataset. |
---|
| 118 | * |
---|
| 119 | * @param instances The dataset. |
---|
| 120 | * @param repeatPredicate True if predicates can be repeated. |
---|
| 121 | * @param maxLiterals Maximum number of literals. |
---|
| 122 | * @param negBody True if negation is allowed in the body. |
---|
| 123 | * @param negHead True if negation is allowed in the head. |
---|
| 124 | * @param classRule True if the rule is a classification rule. |
---|
| 125 | * @param horn True if the rule is a horn clause. |
---|
| 126 | */ |
---|
| 127 | public Rule(Instances instances, |
---|
| 128 | boolean repeatPredicate, int maxLiterals, boolean negBody, |
---|
| 129 | boolean negHead, boolean classRule, boolean horn) { |
---|
| 130 | |
---|
| 131 | m_body = new Body(instances); |
---|
| 132 | m_head = new Head(instances); |
---|
| 133 | m_repeatPredicate = repeatPredicate; |
---|
| 134 | m_maxLiterals = maxLiterals; |
---|
| 135 | m_negBody = negBody && !horn; |
---|
| 136 | m_negHead = negHead && !horn; |
---|
| 137 | m_classRule = classRule; |
---|
| 138 | m_singleHead = classRule || horn; |
---|
| 139 | m_numInstances = instances.numInstances(); |
---|
| 140 | m_counterInstances = new ArrayList(m_numInstances); |
---|
| 141 | Enumeration enu = instances.enumerateInstances(); |
---|
| 142 | while (enu.hasMoreElements()) { |
---|
| 143 | m_counterInstances.add(enu.nextElement()); |
---|
| 144 | } |
---|
| 145 | } |
---|
| 146 | |
---|
| 147 | /** |
---|
| 148 | * Returns a shallow copy of this rule. |
---|
| 149 | * The structured is copied but the literals themselves are not copied. |
---|
| 150 | * |
---|
| 151 | * @return A copy of this Rule. |
---|
| 152 | */ |
---|
| 153 | public Object clone() { |
---|
| 154 | |
---|
| 155 | Object result = null; |
---|
| 156 | try { |
---|
| 157 | result = super.clone(); |
---|
| 158 | /* Clone the body and the head. */ |
---|
| 159 | ((Rule) result).m_body = (Body) m_body.clone(); |
---|
| 160 | ((Rule) result).m_head = (Head) m_head.clone(); |
---|
| 161 | /* Clone the set of counter-instances. */ |
---|
| 162 | if (m_counterInstances != null) { |
---|
| 163 | ((Rule) result).m_counterInstances |
---|
| 164 | = (ArrayList) m_counterInstances.clone(); |
---|
| 165 | } |
---|
| 166 | } catch (Exception e) { |
---|
| 167 | /* An exception is not supposed to happen here. */ |
---|
| 168 | e.printStackTrace(); |
---|
| 169 | System.exit(0); |
---|
| 170 | } |
---|
| 171 | return result; |
---|
| 172 | } |
---|
| 173 | |
---|
| 174 | /** |
---|
| 175 | * Test if an instance is a counter-instance of this rule. |
---|
| 176 | * |
---|
| 177 | * @param instance The instance to test. |
---|
| 178 | * @return True if the instance is a counter-instance. |
---|
| 179 | */ |
---|
| 180 | public boolean counterInstance(Instance instance) { |
---|
| 181 | |
---|
| 182 | return ((m_body.counterInstance(instance) |
---|
| 183 | && m_head.counterInstance(instance))); |
---|
| 184 | } |
---|
| 185 | |
---|
| 186 | /** |
---|
| 187 | * Update the number of counter-instances of this rule in the dataset. |
---|
| 188 | * This method should be used is the rule does not store its counter-instances. |
---|
| 189 | * |
---|
| 190 | * @param instances The dataset. |
---|
| 191 | */ |
---|
| 192 | public void upDate(Instances instances) { |
---|
| 193 | |
---|
| 194 | Enumeration enu = instances.enumerateInstances(); |
---|
| 195 | m_numInstances = instances.numInstances(); |
---|
| 196 | m_counter = 0; |
---|
| 197 | while (enu.hasMoreElements()) { |
---|
| 198 | if (this.counterInstance((Instance) enu.nextElement())) { |
---|
| 199 | m_counter++; |
---|
| 200 | } |
---|
| 201 | } |
---|
| 202 | m_head.upDate(instances); |
---|
| 203 | m_body.upDate(instances); |
---|
| 204 | } |
---|
| 205 | |
---|
| 206 | /** |
---|
| 207 | * Get the confirmation value of this rule. |
---|
| 208 | * |
---|
| 209 | * @return The confirmation. |
---|
| 210 | */ |
---|
| 211 | public double getConfirmation() { |
---|
| 212 | |
---|
| 213 | return m_confirmation; |
---|
| 214 | } |
---|
| 215 | |
---|
| 216 | /** |
---|
| 217 | * Get the optimistic estimate of the confirmation obtained by refining |
---|
| 218 | * this rule. |
---|
| 219 | * |
---|
| 220 | * @return The optimistic estimate. |
---|
| 221 | */ |
---|
| 222 | public double getOptimistic() { |
---|
| 223 | |
---|
| 224 | return m_optimistic; |
---|
| 225 | } |
---|
| 226 | |
---|
| 227 | /* |
---|
| 228 | * Get the expected number of counter-instances of this rule, |
---|
| 229 | * calculated from the number of instances satisfying the body and |
---|
| 230 | * the number of instances satisfying the negation of the head. |
---|
| 231 | * |
---|
| 232 | * @return The expected number of counter-instances. |
---|
| 233 | */ |
---|
| 234 | public double getExpectedNumber() { |
---|
| 235 | |
---|
| 236 | return (double) m_body.getCounterInstancesNumber() |
---|
| 237 | * (double) m_head.getCounterInstancesNumber() |
---|
| 238 | / (double) m_numInstances; |
---|
| 239 | } |
---|
| 240 | |
---|
| 241 | /** |
---|
| 242 | * Get the expected frequency of counter-instances of this rule. |
---|
| 243 | * |
---|
| 244 | * @return The expected frequency of counter-instances. |
---|
| 245 | */ |
---|
| 246 | public double getExpectedFrequency() { |
---|
| 247 | |
---|
| 248 | return getExpectedNumber() / (double) m_numInstances; |
---|
| 249 | } |
---|
| 250 | |
---|
| 251 | /** |
---|
| 252 | * Get the observed number of counter-instances of this rule in the dataset. |
---|
| 253 | * |
---|
| 254 | * @return The observed number of counter-instances. |
---|
| 255 | */ |
---|
| 256 | public int getObservedNumber() { |
---|
| 257 | |
---|
| 258 | if (m_counterInstances != null) { |
---|
| 259 | return m_counterInstances.size(); |
---|
| 260 | } else { |
---|
| 261 | return m_counter; |
---|
| 262 | } |
---|
| 263 | } |
---|
| 264 | |
---|
| 265 | /** |
---|
| 266 | * Get the observed frequency of counter-instances of this rule in the dataset. |
---|
| 267 | * |
---|
| 268 | * @return The expected frequency of counter-instances. |
---|
| 269 | */ |
---|
| 270 | public double getObservedFrequency() { |
---|
| 271 | |
---|
| 272 | return (double) getObservedNumber() / (double) m_numInstances; |
---|
| 273 | } |
---|
| 274 | |
---|
| 275 | /** |
---|
| 276 | * Get the rate of True Positive instances of this rule. |
---|
| 277 | * |
---|
| 278 | * @return The TP-rate. |
---|
| 279 | */ |
---|
| 280 | public double getTPRate() { |
---|
| 281 | |
---|
| 282 | int tp = m_body.getCounterInstancesNumber() - getObservedNumber(); |
---|
| 283 | int fn = m_numInstances - m_head.getCounterInstancesNumber() - tp; |
---|
| 284 | return ((double) tp / (double) (tp + fn)); |
---|
| 285 | } |
---|
| 286 | |
---|
| 287 | /** |
---|
| 288 | * Get the rate of False Positive instances of this rule. |
---|
| 289 | * |
---|
| 290 | * @return The FP-rate. |
---|
| 291 | */ |
---|
| 292 | public double getFPRate() { |
---|
| 293 | |
---|
| 294 | int fp = getObservedNumber(); |
---|
| 295 | int tn = m_head.getCounterInstancesNumber() - fp; |
---|
| 296 | return ((double) fp / (double) (fp + tn)); |
---|
| 297 | } |
---|
| 298 | |
---|
| 299 | /** |
---|
| 300 | * Calculate the confirmation of this rule. |
---|
| 301 | */ |
---|
| 302 | public void calculateConfirmation() { |
---|
| 303 | |
---|
| 304 | double expected = getExpectedFrequency(); |
---|
| 305 | double observed = getObservedFrequency(); |
---|
| 306 | if ((expected == 0) || (expected == 1)) { |
---|
| 307 | m_confirmation = 0; |
---|
| 308 | } else { |
---|
| 309 | m_confirmation = (expected - observed) / (Math.sqrt(expected) - expected); |
---|
| 310 | } |
---|
| 311 | } |
---|
| 312 | |
---|
| 313 | /** |
---|
| 314 | * Calculate the optimistic estimate of this rule. |
---|
| 315 | */ |
---|
| 316 | public void calculateOptimistic() { |
---|
| 317 | |
---|
| 318 | int counterInstances = this.getObservedNumber(); |
---|
| 319 | int body = m_body.getCounterInstancesNumber(); |
---|
| 320 | int notHead = m_head.getCounterInstancesNumber(); |
---|
| 321 | int n = m_numInstances; |
---|
| 322 | double expectedOptimistic; |
---|
| 323 | /* optimistic expected number of counter-instances */ |
---|
| 324 | if (counterInstances <= body - notHead) { |
---|
| 325 | expectedOptimistic = (double) (notHead * (body - counterInstances)) |
---|
| 326 | / (double) (n * n); |
---|
| 327 | } else if (counterInstances <= notHead - body) { |
---|
| 328 | expectedOptimistic = (double) (body * (notHead - counterInstances)) |
---|
| 329 | / (double) (n * n); |
---|
| 330 | } else { |
---|
| 331 | expectedOptimistic = (double) ((notHead + body - counterInstances) |
---|
| 332 | * (notHead + body - counterInstances)) |
---|
| 333 | / (double) (4 * n * n); |
---|
| 334 | } |
---|
| 335 | if ((expectedOptimistic == 0) || (expectedOptimistic == 1)) { |
---|
| 336 | m_optimistic = 0; |
---|
| 337 | } else { |
---|
| 338 | m_optimistic = expectedOptimistic |
---|
| 339 | / (Math.sqrt(expectedOptimistic) - expectedOptimistic); |
---|
| 340 | } |
---|
| 341 | } |
---|
| 342 | |
---|
| 343 | /** |
---|
| 344 | * Test if this rule is empty. |
---|
| 345 | * |
---|
| 346 | * @return True if it is the empty rule. |
---|
| 347 | */ |
---|
| 348 | public boolean isEmpty() { |
---|
| 349 | |
---|
| 350 | return (m_head.isEmpty() && m_body.isEmpty()); |
---|
| 351 | } |
---|
| 352 | |
---|
| 353 | /** |
---|
| 354 | * Give the number of literals in this rule. |
---|
| 355 | * |
---|
| 356 | * @return The number of literals. |
---|
| 357 | */ |
---|
| 358 | public int numLiterals() { |
---|
| 359 | |
---|
| 360 | return m_body.numLiterals() + m_head.numLiterals(); |
---|
| 361 | } |
---|
| 362 | |
---|
| 363 | /** |
---|
| 364 | * Add a literal to the body of the rule. |
---|
| 365 | * |
---|
| 366 | * @param newLit The literal to add. |
---|
| 367 | * @return The rule obtained by adding the literal, null if the literal can |
---|
| 368 | * not be added because of the constraints on the rule. |
---|
| 369 | */ |
---|
| 370 | private Rule addTermToBody(Literal newLit) { |
---|
| 371 | |
---|
| 372 | if (!m_negBody && newLit.negative() |
---|
| 373 | || (m_classRule && newLit.getPredicate().isClass()) |
---|
| 374 | || (newLit instanceof IndividualLiteral |
---|
| 375 | && (((IndividualLiteral) newLit).getType() |
---|
| 376 | - m_body.getType()) > 1 |
---|
| 377 | && (((IndividualLiteral) newLit).getType() |
---|
| 378 | - m_head.getType()) > 1)) { |
---|
| 379 | return null; |
---|
| 380 | } else { |
---|
| 381 | Rule result = (Rule) this.clone(); |
---|
| 382 | result.m_body.addElement(newLit); |
---|
| 383 | /* Update the counter-instances. */ |
---|
| 384 | if (m_counterInstances != null) { |
---|
| 385 | for (int i = result.m_counterInstances.size() - 1; i >= 0; i--) { |
---|
| 386 | Instance current = (Instance) result.m_counterInstances.get(i); |
---|
| 387 | if (!result.m_body.canKeep(current, newLit)) { |
---|
| 388 | result.m_counterInstances.remove(i); |
---|
| 389 | } |
---|
| 390 | } |
---|
| 391 | } |
---|
| 392 | return result; |
---|
| 393 | } |
---|
| 394 | } |
---|
| 395 | |
---|
| 396 | /** |
---|
| 397 | * Add a literal to the head of the rule. |
---|
| 398 | * |
---|
| 399 | * @param newLit The literal to add. |
---|
| 400 | * @return The rule obtained by adding the literal, null if the literal can |
---|
| 401 | * not be added because of the constraints on the rule. |
---|
| 402 | */ |
---|
| 403 | private Rule addTermToHead(Literal newLit) { |
---|
| 404 | if ((!m_negHead && newLit.negative()) |
---|
| 405 | || (m_classRule && !newLit.getPredicate().isClass()) |
---|
| 406 | || (m_singleHead && !m_head.isEmpty()) |
---|
| 407 | || (newLit instanceof IndividualLiteral |
---|
| 408 | && ((IndividualLiteral) newLit).getType() |
---|
| 409 | != IndividualLiteral.INDIVIDUAL_PROPERTY)) { |
---|
| 410 | return null; |
---|
| 411 | } else { |
---|
| 412 | Rule result = (Rule) this.clone(); |
---|
| 413 | result.m_head.addElement(newLit); |
---|
| 414 | /* Update counter-instances. */ |
---|
| 415 | if (m_counterInstances != null) { |
---|
| 416 | for (int i = result.m_counterInstances.size() - 1; i >= 0; i--) { |
---|
| 417 | Instance current = (Instance) result.m_counterInstances.get(i); |
---|
| 418 | if (!result.m_head.canKeep(current, newLit)) { |
---|
| 419 | result.m_counterInstances.remove(i); |
---|
| 420 | } |
---|
| 421 | } |
---|
| 422 | } |
---|
| 423 | return result; |
---|
| 424 | } |
---|
| 425 | } |
---|
| 426 | |
---|
| 427 | /** |
---|
| 428 | * Refine a rule by adding a range of literals of a predicate, either to |
---|
| 429 | * the head or to the body of the rule. |
---|
| 430 | * |
---|
| 431 | * @param pred The predicate to consider. |
---|
| 432 | * @param firstIndex The index of the first literal of the predicate to add. |
---|
| 433 | * @param lastIndex The index of the last literal of the predicate to add. |
---|
| 434 | * @param addTobody True if the literals should be added to the body. |
---|
| 435 | * @param addToHead True if the literals should be added to the head. |
---|
| 436 | * @return A list of rules obtained by refinement. |
---|
| 437 | */ |
---|
| 438 | private SimpleLinkedList refine(Predicate pred, int firstIndex, int lastIndex, |
---|
| 439 | boolean addToBody, boolean addToHead) { |
---|
| 440 | |
---|
| 441 | SimpleLinkedList result = new SimpleLinkedList(); |
---|
| 442 | Literal currentLit; |
---|
| 443 | Literal negation; |
---|
| 444 | Rule refinement; |
---|
| 445 | for (int i = firstIndex; i < lastIndex; i++) { |
---|
| 446 | currentLit = pred.getLiteral(i); |
---|
| 447 | if (addToBody) { |
---|
| 448 | refinement = addTermToBody(currentLit); |
---|
| 449 | if (refinement != null) { |
---|
| 450 | result.add(refinement); |
---|
| 451 | } |
---|
| 452 | } |
---|
| 453 | if (addToHead) { |
---|
| 454 | refinement = addTermToHead(currentLit); |
---|
| 455 | if (refinement != null) { |
---|
| 456 | result.add(refinement); |
---|
| 457 | } |
---|
| 458 | } |
---|
| 459 | negation = currentLit.getNegation(); |
---|
| 460 | if (negation != null) { |
---|
| 461 | if (addToBody) { |
---|
| 462 | refinement = addTermToBody(negation); |
---|
| 463 | if (refinement != null) { |
---|
| 464 | result.add(refinement); |
---|
| 465 | } |
---|
| 466 | } |
---|
| 467 | if (addToHead) { |
---|
| 468 | refinement = addTermToHead(negation); |
---|
| 469 | if (refinement != null) { |
---|
| 470 | result.add(refinement); |
---|
| 471 | } |
---|
| 472 | } |
---|
| 473 | } |
---|
| 474 | } |
---|
| 475 | return result; |
---|
| 476 | } |
---|
| 477 | |
---|
| 478 | /** |
---|
| 479 | * Refine a rule by adding literal from a set of predictes. |
---|
| 480 | * |
---|
| 481 | * @param predicates The predicates available. |
---|
| 482 | * @return The list of the children obtained by refining the rule. |
---|
| 483 | */ |
---|
| 484 | public SimpleLinkedList refine(ArrayList predicates) { |
---|
| 485 | |
---|
| 486 | SimpleLinkedList result = new SimpleLinkedList(); |
---|
| 487 | Predicate currentPred; |
---|
| 488 | boolean addToBody; |
---|
| 489 | boolean addToHead; |
---|
| 490 | |
---|
| 491 | if (this.numLiterals() == m_maxLiterals) { |
---|
| 492 | return result; |
---|
| 493 | } |
---|
| 494 | |
---|
| 495 | if (this.isEmpty()) { |
---|
| 496 | /* Literals can be added on both sides of the rule. */ |
---|
| 497 | for (int i = 0; i < predicates.size(); i++) { |
---|
| 498 | currentPred = (Predicate) predicates.get(i); |
---|
| 499 | result.addAll(refine(currentPred, |
---|
| 500 | 0, currentPred.numLiterals(), |
---|
| 501 | true, true)); |
---|
| 502 | } |
---|
| 503 | } else if (m_body.isEmpty() || m_head.isEmpty()) { |
---|
| 504 | /* Literals can be added to the empty side only. */ |
---|
| 505 | LiteralSet side; |
---|
| 506 | Literal last; |
---|
| 507 | if (m_body.isEmpty()) { |
---|
| 508 | side = m_head; |
---|
| 509 | addToBody = true; |
---|
| 510 | addToHead = false; |
---|
| 511 | } else { // m_head.isEmpty() |
---|
| 512 | side = m_body; |
---|
| 513 | addToBody = false; |
---|
| 514 | addToHead = true; |
---|
| 515 | } |
---|
| 516 | last = side.getLastLiteral(); |
---|
| 517 | currentPred = last.getPredicate(); |
---|
| 518 | if (m_repeatPredicate) { |
---|
| 519 | result.addAll(refine(currentPred, |
---|
| 520 | currentPred.indexOf(last) + 1, |
---|
| 521 | currentPred.numLiterals(), |
---|
| 522 | addToBody, addToHead)); |
---|
| 523 | } |
---|
| 524 | for (int i = predicates.indexOf(currentPred) + 1; i < predicates.size(); |
---|
| 525 | i++) { |
---|
| 526 | currentPred = (Predicate) predicates.get(i); |
---|
| 527 | result.addAll(refine(currentPred, |
---|
| 528 | 0, currentPred.numLiterals(), |
---|
| 529 | addToBody, addToHead)); |
---|
| 530 | } |
---|
| 531 | } else { |
---|
| 532 | Literal lastLitBody = m_body.getLastLiteral(); |
---|
| 533 | Literal lastLitHead = m_head.getLastLiteral(); |
---|
| 534 | Predicate lastPredBody = lastLitBody.getPredicate(); |
---|
| 535 | Predicate lastPredHead = lastLitHead.getPredicate(); |
---|
| 536 | int lastLitBodyIndex = lastPredBody.indexOf(lastLitBody); |
---|
| 537 | int lastLitHeadIndex = lastPredHead.indexOf(lastLitHead); |
---|
| 538 | int lastPredBodyIndex = predicates.indexOf(lastPredBody); |
---|
| 539 | int lastPredHeadIndex = predicates.indexOf(lastPredHead); |
---|
| 540 | Predicate inferiorPred; |
---|
| 541 | Predicate superiorPred; |
---|
| 542 | int inferiorLit; |
---|
| 543 | int superiorLit; |
---|
| 544 | addToBody = (m_head.numLiterals() == 1 |
---|
| 545 | && (lastPredBodyIndex < lastPredHeadIndex |
---|
| 546 | || (lastPredBodyIndex == lastPredHeadIndex |
---|
| 547 | && lastLitBodyIndex < lastLitHeadIndex))); |
---|
| 548 | addToHead = (m_body.numLiterals() == 1 |
---|
| 549 | && (lastPredHeadIndex < lastPredBodyIndex |
---|
| 550 | || (lastPredHeadIndex == lastPredBodyIndex |
---|
| 551 | && lastLitHeadIndex < lastLitBodyIndex))); |
---|
| 552 | if (addToBody || addToHead) { |
---|
| 553 | /* Add literals in the gap between the body and the head. */ |
---|
| 554 | if (addToBody) { |
---|
| 555 | inferiorPred = lastPredBody; |
---|
| 556 | inferiorLit = lastLitBodyIndex; |
---|
| 557 | superiorPred = lastPredHead; |
---|
| 558 | superiorLit = lastLitHeadIndex; |
---|
| 559 | } else { // addToHead |
---|
| 560 | inferiorPred = lastPredHead; |
---|
| 561 | inferiorLit = lastLitHeadIndex; |
---|
| 562 | superiorPred = lastPredBody; |
---|
| 563 | superiorLit = lastLitBodyIndex; |
---|
| 564 | } |
---|
| 565 | if (predicates.indexOf(inferiorPred) |
---|
| 566 | < predicates.indexOf(superiorPred)) { |
---|
| 567 | if (m_repeatPredicate) { |
---|
| 568 | result.addAll(refine(inferiorPred, |
---|
| 569 | inferiorLit + 1, inferiorPred.numLiterals(), |
---|
| 570 | addToBody, addToHead)); |
---|
| 571 | } |
---|
| 572 | for (int j = predicates.indexOf(inferiorPred) + 1; |
---|
| 573 | j < predicates.indexOf(superiorPred); j++) { |
---|
| 574 | currentPred = (Predicate) predicates.get(j); |
---|
| 575 | result.addAll(refine(currentPred, |
---|
| 576 | 0, currentPred.numLiterals(), |
---|
| 577 | addToBody, addToHead)); |
---|
| 578 | } |
---|
| 579 | if (m_repeatPredicate) { |
---|
| 580 | result.addAll(refine(superiorPred, |
---|
| 581 | 0, superiorLit, |
---|
| 582 | addToBody, addToHead)); |
---|
| 583 | } |
---|
| 584 | } else { |
---|
| 585 | //((inferiorPred.getIndex() == superiorPred.getIndex()) |
---|
| 586 | //&& (inferiorLit < superiorLit)) |
---|
| 587 | if (m_repeatPredicate) { |
---|
| 588 | result.addAll(refine(inferiorPred, |
---|
| 589 | inferiorLit + 1, superiorLit, |
---|
| 590 | addToBody, addToHead)); |
---|
| 591 | } |
---|
| 592 | } |
---|
| 593 | } |
---|
| 594 | /* Add other literals. */ |
---|
| 595 | if (predicates.indexOf(lastPredBody) > predicates.indexOf(lastPredHead)) { |
---|
| 596 | superiorPred = lastPredBody; |
---|
| 597 | superiorLit = lastPredBody.indexOf(lastLitBody); |
---|
| 598 | } else if (predicates.indexOf(lastPredBody) |
---|
| 599 | < predicates.indexOf(lastPredHead)) { |
---|
| 600 | superiorPred = lastPredHead; |
---|
| 601 | superiorLit = lastPredHead.indexOf(lastLitHead); |
---|
| 602 | } else { |
---|
| 603 | superiorPred = lastPredBody; |
---|
| 604 | if (lastLitBodyIndex > lastLitHeadIndex) { |
---|
| 605 | superiorLit = lastPredBody.indexOf(lastLitBody); |
---|
| 606 | } else { |
---|
| 607 | superiorLit = lastPredHead.indexOf(lastLitHead); |
---|
| 608 | } |
---|
| 609 | } |
---|
| 610 | if (m_repeatPredicate) { |
---|
| 611 | result.addAll(refine(superiorPred, |
---|
| 612 | superiorLit + 1, superiorPred.numLiterals(), |
---|
| 613 | true, true)); |
---|
| 614 | } |
---|
| 615 | for (int j = predicates.indexOf(superiorPred) + 1; j < predicates.size(); |
---|
| 616 | j++) { |
---|
| 617 | currentPred = (Predicate) predicates.get(j); |
---|
| 618 | result.addAll(refine(currentPred, |
---|
| 619 | 0, currentPred.numLiterals(), |
---|
| 620 | true, true)); |
---|
| 621 | } |
---|
| 622 | } |
---|
| 623 | return result; |
---|
| 624 | } |
---|
| 625 | |
---|
| 626 | /** |
---|
| 627 | * Test if this rule subsumes another rule. |
---|
| 628 | * |
---|
| 629 | * @param otherRule The other rule. |
---|
| 630 | * @return True if the other rule is subsumed. |
---|
| 631 | */ |
---|
| 632 | public boolean subsumes(Rule otherRule) { |
---|
| 633 | |
---|
| 634 | if (this.numLiterals() > otherRule.numLiterals()) { |
---|
| 635 | return false; |
---|
| 636 | } |
---|
| 637 | return (m_body.isIncludedIn(otherRule) && m_head.isIncludedIn(otherRule)); |
---|
| 638 | } |
---|
| 639 | |
---|
| 640 | /** |
---|
| 641 | * Test if this rule and another rule correspond to the same clause. |
---|
| 642 | * |
---|
| 643 | * @param otherRule The other rule. |
---|
| 644 | * @return True if both rules correspond to the same clause. |
---|
| 645 | */ |
---|
| 646 | public boolean sameClauseAs(Rule otherRule) { |
---|
| 647 | |
---|
| 648 | return (this.numLiterals() == otherRule.numLiterals() |
---|
| 649 | && this.subsumes(otherRule)); |
---|
| 650 | } |
---|
| 651 | |
---|
| 652 | /** |
---|
| 653 | * Test if this rule is equivalent to another rule. |
---|
| 654 | * |
---|
| 655 | * @param otherRule The other rule. |
---|
| 656 | * @return True if both rules are equivalent. |
---|
| 657 | */ |
---|
| 658 | public boolean equivalentTo(Rule otherRule) { |
---|
| 659 | |
---|
| 660 | return (this.numLiterals() == otherRule.numLiterals() |
---|
| 661 | && m_head.negationIncludedIn(otherRule.m_body) |
---|
| 662 | && m_body.negationIncludedIn(otherRule.m_head)); |
---|
| 663 | } |
---|
| 664 | |
---|
| 665 | /** |
---|
| 666 | * Test if the body of the rule contains a literal. |
---|
| 667 | * |
---|
| 668 | * @param lit The literal to look for. |
---|
| 669 | * @return True if the literal is contained in the body of the rule. |
---|
| 670 | */ |
---|
| 671 | public boolean bodyContains(Literal lit) { |
---|
| 672 | |
---|
| 673 | return m_body.contains(lit); |
---|
| 674 | } |
---|
| 675 | |
---|
| 676 | /** |
---|
| 677 | * Test if the head of the rule contains a literal. |
---|
| 678 | * |
---|
| 679 | * @param lit The literal to look for. |
---|
| 680 | * @return True if the literal is contained in the head of the rule. |
---|
| 681 | */ |
---|
| 682 | public boolean headContains(Literal lit) { |
---|
| 683 | |
---|
| 684 | return m_head.contains(lit); |
---|
| 685 | } |
---|
| 686 | |
---|
| 687 | /** |
---|
| 688 | * Test if this rule is over the frequency threshold. |
---|
| 689 | * |
---|
| 690 | * @param minFrequency The frequency threshold. |
---|
| 691 | * @return True if the rule is over the threshold. |
---|
| 692 | */ |
---|
| 693 | public boolean overFrequencyThreshold(double minFrequency) { |
---|
| 694 | |
---|
| 695 | return (m_body.overFrequencyThreshold(minFrequency) |
---|
| 696 | && m_head.overFrequencyThreshold(minFrequency)); |
---|
| 697 | } |
---|
| 698 | |
---|
| 699 | /** |
---|
| 700 | * Test if the body of the rule is true. |
---|
| 701 | * |
---|
| 702 | * @return True if the body is always satisfied. |
---|
| 703 | */ |
---|
| 704 | public boolean hasTrueBody() { |
---|
| 705 | |
---|
| 706 | return (!m_body.isEmpty() |
---|
| 707 | && m_body.hasMaxCounterInstances()); |
---|
| 708 | } |
---|
| 709 | |
---|
| 710 | /** |
---|
| 711 | * Test if the head of the rule is false. |
---|
| 712 | * |
---|
| 713 | * @return True if the body is never satisfied. |
---|
| 714 | */ |
---|
| 715 | public boolean hasFalseHead() { |
---|
| 716 | |
---|
| 717 | return (!m_head.isEmpty() |
---|
| 718 | && m_head.hasMaxCounterInstances()); |
---|
| 719 | } |
---|
| 720 | |
---|
| 721 | /** |
---|
| 722 | * Return a String giving the confirmation and optimistic estimate of |
---|
| 723 | * this rule. |
---|
| 724 | * |
---|
| 725 | * @return A String with the values of the rule. |
---|
| 726 | */ |
---|
| 727 | public String valuesToString() { |
---|
| 728 | |
---|
| 729 | StringBuffer text = new StringBuffer(); |
---|
| 730 | DecimalFormat decimalFormat = new DecimalFormat("0.000000"); |
---|
| 731 | text.append(decimalFormat.format(getConfirmation())); |
---|
| 732 | text.append(" "); |
---|
| 733 | text.append(decimalFormat.format(getObservedFrequency())); |
---|
| 734 | return text.toString(); |
---|
| 735 | } |
---|
| 736 | |
---|
| 737 | /** |
---|
| 738 | * Return a String giving the TP-rate and FP-rate of |
---|
| 739 | * this rule. |
---|
| 740 | * |
---|
| 741 | * @return A String with the values of the rule. |
---|
| 742 | */ |
---|
| 743 | public String rocToString() { |
---|
| 744 | |
---|
| 745 | StringBuffer text = new StringBuffer(); |
---|
| 746 | DecimalFormat decimalFormat = new DecimalFormat("0.000000"); |
---|
| 747 | text.append(decimalFormat.format(getConfirmation())); |
---|
| 748 | text.append(" "); |
---|
| 749 | text.append(decimalFormat.format(getTPRate())); |
---|
| 750 | text.append(" "); |
---|
| 751 | text.append(decimalFormat.format(getFPRate())); |
---|
| 752 | return text.toString(); |
---|
| 753 | } |
---|
| 754 | |
---|
| 755 | /** |
---|
| 756 | * Retrun a String for this rule. |
---|
| 757 | * |
---|
| 758 | * @return The String describing this rule. |
---|
| 759 | */ |
---|
| 760 | public String toString() { |
---|
| 761 | |
---|
| 762 | StringBuffer text = new StringBuffer(); |
---|
| 763 | text.append(m_body.toString()); |
---|
| 764 | text.append(" ==> "); |
---|
| 765 | text.append(m_head.toString()); |
---|
| 766 | return text.toString(); |
---|
| 767 | } |
---|
| 768 | |
---|
| 769 | /** |
---|
| 770 | * Comparator used to compare two rules according to their confirmation value. |
---|
| 771 | */ |
---|
| 772 | public static Comparator confirmationComparator = new Comparator() { |
---|
| 773 | |
---|
| 774 | public int compare(Object o1, Object o2) { |
---|
| 775 | |
---|
| 776 | Rule r1 = (Rule) o1; |
---|
| 777 | Rule r2 = (Rule) o2; |
---|
| 778 | double conf1 = r1.getConfirmation(); |
---|
| 779 | double conf2 = r2.getConfirmation(); |
---|
| 780 | if (conf1 > conf2) { |
---|
| 781 | return -1; |
---|
| 782 | } else if (conf1 < conf2) { |
---|
| 783 | return 1; |
---|
| 784 | } else { |
---|
| 785 | return 0; |
---|
| 786 | } |
---|
| 787 | } |
---|
| 788 | }; |
---|
| 789 | |
---|
| 790 | /** |
---|
| 791 | * Comparator used to compare two rules according to their observed number |
---|
| 792 | * of counter-instances. |
---|
| 793 | */ |
---|
| 794 | public static Comparator observedComparator = new Comparator() { |
---|
| 795 | |
---|
| 796 | public int compare(Object o1, Object o2) { |
---|
| 797 | |
---|
| 798 | Rule r1 = (Rule) o1; |
---|
| 799 | Rule r2 = (Rule) o2; |
---|
| 800 | double obs1 = r1.getObservedFrequency(); |
---|
| 801 | double obs2 = r2.getObservedFrequency(); |
---|
| 802 | if (obs1 < obs2) { |
---|
| 803 | return -1; |
---|
| 804 | } else if (obs1 > obs2) { |
---|
| 805 | return 1; |
---|
| 806 | } else { |
---|
| 807 | return 0; |
---|
| 808 | } |
---|
| 809 | } |
---|
| 810 | }; |
---|
| 811 | |
---|
| 812 | /** |
---|
| 813 | * Comparator used to compare two rules according to their optimistic estimate. |
---|
| 814 | */ |
---|
| 815 | public static Comparator optimisticComparator = new Comparator() { |
---|
| 816 | |
---|
| 817 | public int compare(Object o1, Object o2) { |
---|
| 818 | |
---|
| 819 | Rule r1 = (Rule) o1; |
---|
| 820 | Rule r2 = (Rule) o2; |
---|
| 821 | double opt1 = r1.getOptimistic(); |
---|
| 822 | double opt2 = r2.getOptimistic(); |
---|
| 823 | if (opt1 > opt2) { |
---|
| 824 | return -1; |
---|
| 825 | } else if (opt1 < opt2) { |
---|
| 826 | return 1; |
---|
| 827 | } else { |
---|
| 828 | return 0; |
---|
| 829 | } |
---|
| 830 | } |
---|
| 831 | }; |
---|
| 832 | |
---|
| 833 | /** |
---|
| 834 | * Comparator used to compare two rules according to their confirmation and |
---|
| 835 | * then their observed number of counter-instances. |
---|
| 836 | */ |
---|
| 837 | public static Comparator confirmationThenObservedComparator = new Comparator() { |
---|
| 838 | public int compare(Object o1, Object o2) { |
---|
| 839 | |
---|
| 840 | int confirmationComparison = confirmationComparator.compare(o1, o2); |
---|
| 841 | if (confirmationComparison != 0) { |
---|
| 842 | return confirmationComparison; |
---|
| 843 | } else { |
---|
| 844 | return observedComparator.compare(o1, o2); |
---|
| 845 | } |
---|
| 846 | } |
---|
| 847 | }; |
---|
| 848 | |
---|
| 849 | /** |
---|
| 850 | * Comparator used to compare two rules according to their optimistic estimate |
---|
| 851 | * and then their observed number of counter-instances. |
---|
| 852 | */ |
---|
| 853 | public static Comparator optimisticThenObservedComparator = new Comparator() { |
---|
| 854 | public int compare(Object o1, Object o2) { |
---|
| 855 | int optimisticComparison = optimisticComparator.compare(o1, o2); |
---|
| 856 | if (optimisticComparison != 0) { |
---|
| 857 | return optimisticComparison; |
---|
| 858 | } else { |
---|
| 859 | return observedComparator.compare(o1, o2); |
---|
| 860 | } |
---|
| 861 | } |
---|
| 862 | }; |
---|
| 863 | |
---|
| 864 | /** |
---|
| 865 | * Returns the revision string. |
---|
| 866 | * |
---|
| 867 | * @return the revision |
---|
| 868 | */ |
---|
| 869 | public String getRevision() { |
---|
| 870 | return RevisionUtils.extract("$Revision: 1.7 $"); |
---|
| 871 | } |
---|
| 872 | } |
---|