1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500
|
package weka.classifiers.rules;
import weka.classifiers.Classifier;
import weka.classifiers.AbstractClassifier;
import weka.core.Attribute;
import weka.core.Capabilities;
import weka.core.Instance;
import weka.core.Instances;
import weka.core.RevisionHandler;
import weka.core.RevisionUtils;
import weka.core.TechnicalInformation;
import weka.core.TechnicalInformationHandler;
import weka.core.Capabilities.Capability;
import weka.core.TechnicalInformation.Field;
import weka.core.TechnicalInformation.Type;
import weka.core.Utils;
import java.io.Serializable;
import java.util.Enumeration;
public class Prism extends AbstractClassifier implements TechnicalInformationHandler {
/** for serialization */
static final long serialVersionUID = 1310258880025902106L;
/**
* Returns a string describing classifier
* @return a description suitable for
* displaying in the explorer/experimenter gui
*/
public String globalInfo() {
return "Class for building and using a PRISM rule set for classification. "
+ "Can only deal with nominal attributes. Can't deal with missing values. "
+ "Doesn't do any pruning.\n\n"
+ "For more information, see \n\n"
+ getTechnicalInformation().toString();
}
/**
* Returns an instance of a TechnicalInformation object, containing
* detailed information about the technical background of this class,
* e.g., paper reference or book this class is based on.
*
* @return the technical information about this class
*/
public TechnicalInformation getTechnicalInformation() {
TechnicalInformation result;
result = new TechnicalInformation(Type.ARTICLE);
result.setValue(Field.AUTHOR, "J. Cendrowska");
result.setValue(Field.YEAR, "1987");
result.setValue(Field.TITLE, "PRISM: An algorithm for inducing modular rules");
result.setValue(Field.JOURNAL, "International Journal of Man-Machine Studies");
result.setValue(Field.VOLUME, "27");
result.setValue(Field.NUMBER, "4");
result.setValue(Field.PAGES, "349-370");
return result;
}
/**
* Class for storing a PRISM ruleset, i.e. a list of rules
*/
private class PrismRule implements Serializable, RevisionHandler {
/** for serialization */
static final long serialVersionUID = 4248784350656508583L;
/** The classification */
private int m_classification;
/** The instance */
private Instances m_instances;
/** First test of this rule */
private Test m_test;
/** Number of errors made by this rule (will end up 0) */
private int m_errors;
/** The next rule in the list */
private PrismRule m_next;
/**
* Constructor that takes instances and the classification.
*
* @param data the instances
* @param cl the class
* @exception Exception if something goes wrong
*/
public PrismRule(Instances data, int cl) throws Exception {
m_instances = data;
m_classification = cl;
m_test = null;
m_next = null;
m_errors = 0;
Enumeration enu = data.enumerateInstances();
while (enu.hasMoreElements()) {
if ((int) ((Instance) enu.nextElement()).classValue() != cl) {
m_errors++;
}
}
m_instances = new Instances(m_instances, 0);
}
/**
* Returns the result assigned by this rule to a given instance.
*
* @param inst the instance to be classified
* @return the classification
*/
public int resultRule(Instance inst) {
if (m_test == null || m_test.satisfies(inst)) {
return m_classification;
} else {
return -1;
}
}
/**
* Returns the result assigned by these rules to a given instance.
*
* @param inst the instance to be classified
* @return the classification
*/
public int resultRules(Instance inst) {
if (resultRule(inst) != -1) {
return m_classification;
} else if (m_next != null) {
return m_next.resultRules(inst);
} else {
return -1;
}
}
/**
* Returns the set of instances that are covered by this rule.
*
* @param data the instances to be checked
* @return the instances covered
*/
public Instances coveredBy(Instances data) {
Instances r = new Instances(data, data.numInstances());
Enumeration enu = data.enumerateInstances();
while (enu.hasMoreElements()) {
Instance i = (Instance) enu.nextElement();
if (resultRule(i) != -1) {
r.add(i);
}
}
r.compactify();
return r;
}
/**
* Returns the set of instances that are not covered by this rule.
*
* @param data the instances to be checked
* @return the instances not covered
*/
public Instances notCoveredBy(Instances data) {
Instances r = new Instances(data, data.numInstances());
Enumeration enu = data.enumerateInstances();
while (enu.hasMoreElements()) {
Instance i = (Instance) enu.nextElement();
if (resultRule(i) == -1) {
r.add(i);
}
}
r.compactify();
return r;
}
/**
* Prints the set of rules.
*
* @return a description of the rules as a string
*/
public String toString() {
try {
StringBuffer text = new StringBuffer();
if (m_test != null) {
text.append("If ");
for (Test t = m_test; t != null; t = t.m_next) {
if (t.m_attr == -1) {
text.append("?");
} else {
text.append(m_instances.attribute(t.m_attr).name() + " = " +
m_instances.attribute(t.m_attr).value(t.m_val));
}
if (t.m_next != null) {
text.append("\n and ");
}
}
text.append(" then ");
}
text.append(m_instances.classAttribute().value(m_classification) + "\n");
if (m_next != null) {
text.append(m_next.toString());
}
return text.toString();
} catch (Exception e) {
return "Can't print Prism classifier!";
}
}
/**
* Returns the revision string.
*
* @return the revision
*/
public String getRevision() {
return RevisionUtils.extract("$Revision: 8109 $");
}
}
/**
* Class for storing a list of attribute-value tests
*/
private class Test
implements Serializable, RevisionHandler {
/** for serialization */
static final long serialVersionUID = -8925333011350280799L;
/** Attribute to test */
private int m_attr = -1;
/** The attribute's value */
private int m_val;
/** The next test in the rule */
private Test m_next = null;
/**
* Returns whether a given instance satisfies this test.
*
* @param inst the instance to be tested
* @return true if the instance satisfies the test
*/
private boolean satisfies(Instance inst) {
if ((int) inst.value(m_attr) == m_val) {
if (m_next == null) {
return true;
} else {
return m_next.satisfies(inst);
}
}
return false;
}
/**
* Returns the revision string.
*
* @return the revision
*/
public String getRevision() {
return RevisionUtils.extract("$Revision: 8109 $");
}
}
/** The first rule in the list of rules */
private PrismRule m_rules;
/**
* Classifies a given instance.
*
* @param inst the instance to be classified
* @return the classification
*/
public double classifyInstance(Instance inst) {
int result = m_rules.resultRules(inst);
if (result == -1) {
return Utils.missingValue();
} else {
return (double)result;
}
}
/**
* Returns default capabilities of the classifier.
*
* @return the capabilities of this classifier
*/
public Capabilities getCapabilities() {
Capabilities result = super.getCapabilities();
result.disableAll();
// attributes
result.enable(Capability.NOMINAL_ATTRIBUTES);
// class
result.enable(Capability.NOMINAL_CLASS);
result.enable(Capability.MISSING_CLASS_VALUES);
return result;
}
/**
* Generates the classifier.
*
* @param data the data to be used
* @exception Exception if the classifier can't built successfully
*/
public void buildClassifier(Instances data) throws Exception {
int cl; // possible value of theClass
Instances E, ruleE;
PrismRule rule = null;
Test test = null, oldTest = null;
int bestCorrect, bestCovers, attUsed;
Enumeration enumAtt;
// can classifier handle the data?
getCapabilities().testWithFail(data);
// remove instances with missing class
data = new Instances(data);
data.deleteWithMissingClass();
for (cl = 0; cl < data.numClasses(); cl++) { // for each class cl
E = data; // initialize E to the instance set
while (contains(E, cl)) { // while E contains examples in class cl
rule = addRule(rule, new PrismRule(E, cl)); // make a new rule
ruleE = E; // examples covered by this rule
while (rule.m_errors != 0) { // until the rule is perfect
test = new Test(); // make a new test
bestCorrect = bestCovers = attUsed = 0;
// for every attribute not mentioned in the rule
enumAtt = ruleE.enumerateAttributes();
while (enumAtt.hasMoreElements()) {
Attribute attr = (Attribute) enumAtt.nextElement();
if (isMentionedIn(attr, rule.m_test)) {
attUsed++;
continue;
}
int M = attr.numValues();
int[] covers = new int [M];
int[] correct = new int [M];
for (int j = 0; j < M; j++) {
covers[j] = correct[j] = 0;
}
// ... calculate the counts for this class
Enumeration enu = ruleE.enumerateInstances();
while (enu.hasMoreElements()) {
Instance i = (Instance) enu.nextElement();
covers[(int) i.value(attr)]++;
if ((int) i.classValue() == cl) {
correct[(int) i.value(attr)]++;
}
}
// ... for each value of this attribute, see if this test is better
for (int val = 0; val < M; val ++) {
int diff = correct[val] * bestCovers - bestCorrect * covers[val];
// this is a ratio test, correct/covers vs best correct/covers
if (test.m_attr == -1
|| diff > 0 || (diff == 0 && correct[val] > bestCorrect)) {
// update the rule to use this test
bestCorrect = correct[val];
bestCovers = covers[val];
test.m_attr = attr.index();
test.m_val = val;
rule.m_errors = bestCovers - bestCorrect;
}
}
}
if (test.m_attr == -1) { // Couldn't find any sensible test
break;
}
oldTest = addTest(rule, oldTest, test);
ruleE = rule.coveredBy(ruleE);
if (attUsed == (data.numAttributes() - 1)) { // Used all attributes.
break;
}
}
E = rule.notCoveredBy(E);
}
}
}
/**
* Add a rule to the ruleset.
*
* @param lastRule the last rule in the rule set
* @param newRule the rule to be added
* @return the new last rule in the rule set
*/
private PrismRule addRule(PrismRule lastRule, PrismRule newRule) {
if (lastRule == null) {
m_rules = newRule;
} else {
lastRule.m_next = newRule;
}
return newRule;
}
/**
* Add a test to this rule.
*
* @param rule the rule to which test is to be added
* @param lastTest the rule's last test
* @param newTest the test to be added
* @return the new last test of the rule
*/
private Test addTest(PrismRule rule, Test lastTest, Test newTest) {
if (rule.m_test == null) {
rule.m_test = newTest;
} else {
lastTest.m_next = newTest;
}
return newTest;
}
/**
* Does E contain any examples in the class C?
*
* @param E the instances to be checked
* @param C the class
* @return true if there are any instances of class C
* @throws Exception if something goes wrong
*/
private static boolean contains(Instances E, int C) throws Exception {
Enumeration enu = E.enumerateInstances();
while (enu.hasMoreElements()) {
if ((int) ((Instance) enu.nextElement()).classValue() == C) {
return true;
}
}
return false;
}
/**
* Is this attribute mentioned in the rule?
*
* @param attr the attribute to be checked for
* @param t test contained by rule
* @return true if the attribute is mentioned in the rule
*/
private static boolean isMentionedIn(Attribute attr, Test t) {
if (t == null) {
return false;
}
if (t.m_attr == attr.index()) {
return true;
}
return isMentionedIn(attr, t.m_next);
}
/**
* Prints a description of the classifier.
*
* @return a description of the classifier as a string
*/
public String toString() {
if (m_rules == null) {
return "Prism: No model built yet.";
}
return "Prism rules\n----------\n" + m_rules.toString();
}
/**
* Returns the revision string.
*
* @return the revision
*/
public String getRevision() {
return RevisionUtils.extract("$Revision: 8109 $");
}
/**
* Main method for testing this class
*
* @param args the commandline parameters
*/
public static void main(String[] args) {
runClassifier(new Prism(), args);
}
} |
Partager