Data Mining: Practical Machine Learning Tools and Techniques, Second Edition
Yüklə 4,3 Mb. Pdf görüntüsü
|
5 0 8 I N D E X causal relations, 350
chain rule (probability theory), 275 character sets, 310
chi-squared test, 302 circular ordering, 51 city-block metric, 129 ClassAssigner, 431 class attribute, 43 class distribution, 304 class hierarchy in Weka, 471–483 classification, 121 Classification And Regression Tree (CART), 29, 38, 199, 253 classification learning, 43 classification problems, 42 classification rules, 65–69, 200–214 converting decision trees to, 198 criteria for choosing tests, 200–201 decision list, 11 different from association rules, 42 global optimization, 205–207 good (worthwhile) rules, 202–205 missing values, 201–202 numeric attributes, 202 partial decision tree, 207–210 pruning, 203, 205 replicated subtree problem, 66–68 RIPPER rule learner, 205 rules with exceptions, 210–213 Weka, 408–409 Classifier, 453 classifier in Weka, 366, 471–483 classifier algorithms. See learning algorithms classifier errors in Weka, 379 ClassifierPerformanceEvaluator, 431 classifiers package, 453–455 ClassifierSubsetEval, 422 Classifier superclass, 480 classifyInstance(), 453, 480–481 Classify panel, 373, 384 classify text files into two categories, 461–469 Classit, 271 class noise, 313 ClassOrder, 403 class summary, 451 ClassValuePicker, 431 cleaning data, 52, 60 automatic, 312 closed world assumption, 45 ClustererPerformanceEvaluator, 431 clustering, 43, 136–139, 254–271 anomalies, 258 basic assumption, 270 Bayesian, 268–270 category utility, 260–262 Cluster panel (Weka), 391–392 document, 353 EM algorithm, 265–266 faster distance calculations, 138–139 hierarchical, 139 how many clusters?, 254–255 incremental, 255–260
MDL principle, 183–184 merging, 257 mixture model, 262–264, 266–268 output, 81–82 probability-based, 262–265 RBF network, 234 splitting, 254–255, 257 unlabeled data, 337–339 clustering algorithms in Weka, 418–419 clustering for classification, 337
Cobweb, 271 Cobweb in Weka, 419 co-EM, 340 column separation, 336 combining multiple models, 287, 315–336 additive logistic regression, 327–328 additive regression, 325–327 bagging, 316–319 bagging with costs, 319–320 boosting, 321–325 error-correcting output codes, 334–336 logistic model trees, 331 option trees, 328–331 P088407-INDEX.qxd 4/30/05 11:25 AM Page 508
I N D E X 5 0 9
randomization, 320–321 stacking, 332–334 command-line interface, 449–459 class, 450, 452 classifiers package, 453–455 core package, 451, 452 generic options, 456–458 instance, 450 Javadoc indices, 456 package, 451, 452 scheme-specific options, 458–459 starting up, 449–450
command-line options, 456–459 comma-separated value (CSV) format, 370, 371
comparing data mining methods, 153–157 ComplementNaiveBayes, 405 compression techniques, 362 computational learning theory, 324
Computer Assisted Passenger Pre-Screening System (CAPPS), 357 computer network security, 357 computer software. See Weka workbench concept, 42 concept description, 42 concept description language, 32 concept representation, 82. See also knowledge representation conditional independence, 275 conditional likelihood for scoring networks, 280, 283 confidence, 69, 113, 324 Confidence, 420 confidence tests, 154–157, 184 conflict resolution strategies, 82 confusion matrix, 163 conjunction, 65
consensus filter, 342 consequent, of rule, 65
constrained quadratic optimization, 217 consumer music, 359 contact lens data, 6, 13–15 continuous attributes, 49. See also numeric attributes continuous monitoring, 28–29 converting discrete to numeric attributes, 304–305 convex hull, 171, 216 Conviction, 420 Copy, 395 core package, 451, 452 corrected resampled t-test, 157 correlation coefficient, 177–179 cost curves, 173–176 cost matrix, 164–165 cost of errors, 161–176 bagging, 319–320 cost curves, 173–176 cost-sensitive classification, 164–165 cost-sensitive learning, 165–166 Kappa statistic, 163–164 lift charts, 166–168 recall-precision curves, 171–172 ROC curves, 168–171 cost-sensitive classification, 164–165
cost-sensitive learning, 165–166 cost-sensitive learning in Weka, 417 co-training, 339–340 covariance matrix, 267, 307 coverage, of association rules, 69 covering algorithm, 106–111 cow culling, 3–4, 37, 161–162 CPU performance data, 16–17 credit approval, 22–23 cross-validated ROC curves, 170 cross-validation, 149–152, 326 inner, 286 outer, 286 repeated, 144
CSV format, 370, 371 P088407-INDEX.qxd 4/30/05 11:25 AM Page 509
Yüklə 4,3 Mb. Dostları ilə paylaş:
|