Data Mining: Practical Machine Learning Tools and Techniques, Second Edition
Yüklə 4,3 Mb. Pdf görüntüsü
|
5 1 2 I N D E X Experimenter, 437–447 advanced panel, 443–445 Analyze panel, 443–445 analyzing the results, 440–441 distributing processing over several machines, 445–447 running an experiment, 439–440 simple setup, 441–442 starting up, 438–441 subexperiments, 447 Explorer, 369–425 ARFF, 370, 371, 380–382 Associate panel, 392 association-rule learners, 419–420 attribute evaluation methods, 421, 422–423
attribute selection, 392–393, 420–425 boosting, 416 classifier errors, 379
clustering algorithms, 418–419 Cluster panel, 391–392 CSV format, 370, 371 error log, 378 file format converters, 380–382 filtering algorithms, 393–403 filters, 382–384 J4.8, 373–377 learning algorithms, 403–414. See also learning algorithms metalearning algorithms, 414–418 models, 377–378 panels, 380 Preprocess panel, 380 search methods, 421, 423–425 Select attributes panel, 392–393 starting up, 369–379 supervised filters, 401–403 training/testing learning schemes, 384–387 unsupervised attribute filters, 395–400 unsupervised instance filters, 400–401 User Classifier, 388–391 Visualize panel, 393 extraction problems, 353, 354 F Fahrenheit, Daniel, 51 fallback heuristic, 239 false negative (FN), 162 false positive (FP), 162 false positive rate, 163 False positive rate, 378 Familiar, 360 family tree, 45 tabular representation of, 46 FarthestFirst, 419 features. See attributes feature selection, 341. See also attribute selection feedforward networks, 233 fielded applications, 22 continuous monitoring, 28–29 customer support and service, 28 cybersecurity, 29 diagnosis, 25–26 ecological applications, 23, 28 electricity supply, 24–25 hazard detection system, 23–24 load forecasting, 24–25 loan application, 22–23 manufacturing processes, 28 marketing and sales, 26–28 oil slick detection, 23 preventive maintenance of electromechanical devices, 25–26 scientific applications, 28 file format converters, 380–382 file mining, 49 filter, 290 filter in Weka, 382–384
filtering algorithms in Weka, 393–403 sparse instances, 401 supervised filters, 401–403 unsupervised attribute filters, 395–400 unsupervised instance filters, 400–401 filtering approaches, 315 filters menu, 383 finite mixture, 262, 263
P088407-INDEX.qxd 4/30/05 11:25 AM Page 512 I N D E X 5 1 3
Fisher, R. A., 15 flat file, 45 F-measure, 172 FN (false negatives), 162 folds, 150 forward pruning, 34, 192 forward selection, 292, 294 forward stagewise additive modeling, 325–327 Fourier analysis, 25 FP (false positives), 162 freedom, degrees of 93, 155 functional dependencies, 350 functions in Weka, 404–405, 409–410
gain ratio, 104 GainRatioAttributeEval, 423 gambling, 160 garbage in, garbage out. See cost of errors; data cleaning; error rate Gaussian-distribution assumption, 92 Gaussian kernel function, 252 generalization as search, 30–35 bias, 32–35 enumerating concept space, 31–32 generalized distance functions, 241–242 generalized exemplars, 236 general-to-specific search bias, 34 genetic algorithms, 38 genetic algorithm search procedures, 294, 341
getting to know your data, 60 global discretization, 297
global optimization, 205–207 Gosset, William, 184 gradient descent, 227, 229, 230 Grading, 417 graphical models, 283 GraphViewer, 431 gray bar in margin of textbook (optional sections), 30 greedy search, 33 GreedyStepwise, 423–424 growing set, 202 H Hamming distance, 335 hand-labeled data, 338 hapax legomena, 310 hard instances, 322 hash table, 280 hazard detection system, 23–24 hidden attributes, 272 hidden layer, 226, 231, 232 hidden units, 226, 231, 234 hierarchical clustering, 139 highly-branching attribute, 86 high-performance rule inducers, 188 histogram equalization, 298 historical literary mystery, 358 holdout method, 146, 149–150, 333 homeland defense, 357 HTML, 355 hypermetrope, 13 hyperpipes, 139 Hyperpipes, 414 hyperplane, 124, 125 hyperrectangle, 238–239 hyperspheres, 133 hypertext markup language (HTML), 355 hypothesis testing, 29 I IB1, 413 IB3, 237
IBk, 413 ID3, 105
Id3, 404 identification code, 86, 102–104 implementation—real-world schemes, 187–283
Bayesian networks, 271–283 classification rules, 200–214 clustering, 254–271 decision tree, 189–199 instance-based, 236–243 linear models, 214–235 P088407-INDEX.qxd 4/30/05 11:25 AM Page 513
Yüklə 4,3 Mb. Dostları ilə paylaş:
|