1 | // Copyright 2015 Georg-August-Universit�t G�ttingen, Germany |
---|
2 | // |
---|
3 | // Licensed under the Apache License, Version 2.0 (the "License"); |
---|
4 | // you may not use this file except in compliance with the License. |
---|
5 | // You may obtain a copy of the License at |
---|
6 | // |
---|
7 | // http://www.apache.org/licenses/LICENSE-2.0 |
---|
8 | // |
---|
9 | // Unless required by applicable law or agreed to in writing, software |
---|
10 | // distributed under the License is distributed on an "AS IS" BASIS, |
---|
11 | // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
---|
12 | // See the License for the specific language governing permissions and |
---|
13 | // limitations under the License. |
---|
14 | |
---|
15 | package de.ugoe.cs.cpdp.training; |
---|
16 | |
---|
17 | import java.util.ArrayList; |
---|
18 | import java.util.Arrays; |
---|
19 | import java.util.Collections; |
---|
20 | import java.util.Comparator; |
---|
21 | import java.util.HashMap; |
---|
22 | import java.util.Iterator; |
---|
23 | import java.util.LinkedHashMap; |
---|
24 | import java.util.LinkedList; |
---|
25 | import java.util.List; |
---|
26 | import java.util.Map; |
---|
27 | import java.util.logging.Level; |
---|
28 | |
---|
29 | import javax.management.RuntimeErrorException; |
---|
30 | |
---|
31 | import java.util.Random; |
---|
32 | |
---|
33 | import org.apache.commons.collections4.list.SetUniqueList; |
---|
34 | import org.apache.commons.math3.stat.inference.ChiSquareTest; |
---|
35 | import org.apache.commons.math3.stat.correlation.SpearmansCorrelation; |
---|
36 | import org.apache.commons.math3.stat.inference.KolmogorovSmirnovTest; |
---|
37 | |
---|
38 | import de.ugoe.cs.util.console.Console; |
---|
39 | import weka.attributeSelection.SignificanceAttributeEval; |
---|
40 | import weka.classifiers.AbstractClassifier; |
---|
41 | import weka.classifiers.Classifier; |
---|
42 | import weka.core.Attribute; |
---|
43 | import weka.core.DenseInstance; |
---|
44 | import weka.core.FastVector; |
---|
45 | import weka.core.Instance; |
---|
46 | import weka.core.Instances; |
---|
47 | |
---|
48 | |
---|
49 | public class MetricMatchingTraining extends WekaBaseTraining implements ISetWiseTestdataAwareTrainingStrategy { |
---|
50 | |
---|
51 | private SetUniqueList<Instances> traindataSet; |
---|
52 | private MetricMatch mm; |
---|
53 | private final Classifier classifier = new MetricMatchingClassifier(); |
---|
54 | |
---|
55 | private String method; |
---|
56 | private float threshold; |
---|
57 | |
---|
58 | /** |
---|
59 | * We wrap the classifier here because of classifyInstance |
---|
60 | * @return |
---|
61 | */ |
---|
62 | @Override |
---|
63 | public Classifier getClassifier() { |
---|
64 | return this.classifier; |
---|
65 | } |
---|
66 | |
---|
67 | |
---|
68 | @Override |
---|
69 | public void setMethod(String method) { |
---|
70 | this.method = method; |
---|
71 | } |
---|
72 | |
---|
73 | |
---|
74 | @Override |
---|
75 | public void setThreshold(String threshold) { |
---|
76 | this.threshold = Float.parseFloat(threshold); |
---|
77 | } |
---|
78 | |
---|
79 | /** |
---|
80 | * We need the testdata instances to do a metric matching, so in this special case we get this data |
---|
81 | * before evaluation |
---|
82 | */ |
---|
83 | @Override |
---|
84 | public void apply(SetUniqueList<Instances> traindataSet, Instances testdata) { |
---|
85 | this.traindataSet = traindataSet; |
---|
86 | |
---|
87 | int rank = 0; // we want at least 5 matching attributes |
---|
88 | int num = 0; |
---|
89 | int biggest_num = 0; |
---|
90 | MetricMatch tmp; |
---|
91 | MetricMatch biggest = null; |
---|
92 | for (Instances traindata : this.traindataSet) { |
---|
93 | num++; |
---|
94 | tmp = new MetricMatch(traindata, testdata); |
---|
95 | |
---|
96 | // metric selection may create error, continue to next training set |
---|
97 | try { |
---|
98 | tmp.attributeSelection(); |
---|
99 | }catch(Exception e) { |
---|
100 | e.printStackTrace(); |
---|
101 | throw new RuntimeException(e); |
---|
102 | } |
---|
103 | |
---|
104 | if (this.method.equals("spearman")) { |
---|
105 | tmp.spearmansRankCorrelation(this.threshold); |
---|
106 | } |
---|
107 | else if (this.method.equals("kolmogorov")) { |
---|
108 | tmp.kolmogorovSmirnovTest(this.threshold); |
---|
109 | } |
---|
110 | else if( this.method.equals("percentile") ) { |
---|
111 | tmp.percentiles(this.threshold); |
---|
112 | } |
---|
113 | else { |
---|
114 | throw new RuntimeException("unknown method"); |
---|
115 | } |
---|
116 | |
---|
117 | // we only select the training data from our set with the most matching attributes |
---|
118 | if (tmp.getRank() > rank) { |
---|
119 | rank = tmp.getRank(); |
---|
120 | biggest = tmp; |
---|
121 | biggest_num = num; |
---|
122 | } |
---|
123 | } |
---|
124 | |
---|
125 | if (biggest == null) { |
---|
126 | throw new RuntimeException("not enough matching attributes found"); |
---|
127 | } |
---|
128 | |
---|
129 | // we use the best match |
---|
130 | |
---|
131 | this.mm = biggest; |
---|
132 | Instances ilist = this.mm.getMatchedTrain(); |
---|
133 | Console.traceln(Level.INFO, "Chosing the trainingdata set num "+biggest_num +" with " + rank + " matching attributes, " + ilist.size() + " instances out of a possible set of " + traindataSet.size() + " sets"); |
---|
134 | |
---|
135 | // replace traindataSEt |
---|
136 | //traindataSet = new SetUniqueList<Instances>(); |
---|
137 | traindataSet.clear(); |
---|
138 | traindataSet.add(ilist); |
---|
139 | |
---|
140 | // we have to build the classifier here: |
---|
141 | try { |
---|
142 | |
---|
143 | // |
---|
144 | if (this.classifier == null) { |
---|
145 | Console.traceln(Level.SEVERE, "Classifier is null"); |
---|
146 | } |
---|
147 | //Console.traceln(Level.INFO, "Building classifier with the matched training data with " + ilist.size() + " instances and "+ ilist.numAttributes() + " attributes"); |
---|
148 | this.classifier.buildClassifier(ilist); |
---|
149 | ((MetricMatchingClassifier) this.classifier).setMetricMatching(this.mm); |
---|
150 | }catch(Exception e) { |
---|
151 | e.printStackTrace(); |
---|
152 | throw new RuntimeException(e); |
---|
153 | } |
---|
154 | } |
---|
155 | |
---|
156 | |
---|
157 | /** |
---|
158 | * encapsulates the classifier configured with WekaBase |
---|
159 | */ |
---|
160 | public class MetricMatchingClassifier extends AbstractClassifier { |
---|
161 | |
---|
162 | private static final long serialVersionUID = -1342172153473770935L; |
---|
163 | private MetricMatch mm; |
---|
164 | private Classifier classifier; |
---|
165 | |
---|
166 | @Override |
---|
167 | public void buildClassifier(Instances traindata) throws Exception { |
---|
168 | this.classifier = setupClassifier(); // parent method from WekaBase |
---|
169 | this.classifier.buildClassifier(traindata); |
---|
170 | } |
---|
171 | |
---|
172 | public void setMetricMatching(MetricMatch mm) { |
---|
173 | this.mm = mm; |
---|
174 | } |
---|
175 | |
---|
176 | /** |
---|
177 | * Here we can not do the metric matching because we only get one instance |
---|
178 | */ |
---|
179 | public double classifyInstance(Instance testdata) { |
---|
180 | // todo: maybe we can pull the instance out of our matched testdata? |
---|
181 | Instance ntest = this.mm.getMatchedTestInstance(testdata); |
---|
182 | |
---|
183 | double ret = 0.0; |
---|
184 | try { |
---|
185 | ret = this.classifier.classifyInstance(ntest); |
---|
186 | }catch(Exception e) { |
---|
187 | e.printStackTrace(); |
---|
188 | throw new RuntimeException(e); |
---|
189 | } |
---|
190 | |
---|
191 | return ret; |
---|
192 | } |
---|
193 | } |
---|
194 | |
---|
195 | /** |
---|
196 | * Encapsulates MetricMatching on Instances Arrays |
---|
197 | */ |
---|
198 | public class MetricMatch { |
---|
199 | Instances train; |
---|
200 | Instances test; |
---|
201 | |
---|
202 | HashMap<Integer, Integer> attributes = new HashMap<Integer,Integer>(); |
---|
203 | |
---|
204 | ArrayList<double[]> train_values; |
---|
205 | ArrayList<double[]> test_values; |
---|
206 | |
---|
207 | // todo: this constructor does not work |
---|
208 | public MetricMatch() { |
---|
209 | } |
---|
210 | |
---|
211 | public MetricMatch(Instances train, Instances test) { |
---|
212 | this.train = new Instances(train); // expensive! but we are dropping the attributes |
---|
213 | this.test = new Instances(test); |
---|
214 | |
---|
215 | // 1. convert metrics of testdata and traindata to later use in test |
---|
216 | this.train_values = new ArrayList<double[]>(); |
---|
217 | for (int i = 0; i < this.train.numAttributes()-1; i++) { |
---|
218 | this.train_values.add(train.attributeToDoubleArray(i)); |
---|
219 | } |
---|
220 | |
---|
221 | this.test_values = new ArrayList<double[]>(); |
---|
222 | for (int i=0; i < this.test.numAttributes()-1; i++) { |
---|
223 | this.test_values.add(this.test.attributeToDoubleArray(i)); |
---|
224 | } |
---|
225 | } |
---|
226 | |
---|
227 | /** |
---|
228 | * returns the number of matched attributes |
---|
229 | * as a way of scoring traindata sets individually |
---|
230 | * |
---|
231 | * @return |
---|
232 | */ |
---|
233 | public int getRank() { |
---|
234 | return this.attributes.size(); |
---|
235 | } |
---|
236 | |
---|
237 | public int getNumInstances() { |
---|
238 | return this.train_values.get(0).length; |
---|
239 | } |
---|
240 | |
---|
241 | public Instance getMatchedTestInstance(Instance test) { |
---|
242 | // create new instance with our matched number of attributes + 1 (the class attribute) |
---|
243 | //Console.traceln(Level.INFO, "getting matched instance"); |
---|
244 | Instances testdata = this.getMatchedTest(); |
---|
245 | |
---|
246 | //Instance ni = new DenseInstance(this.attmatch.size()+1); |
---|
247 | Instance ni = new DenseInstance(this.attributes.size()+1); |
---|
248 | ni.setDataset(testdata); |
---|
249 | |
---|
250 | //Console.traceln(Level.INFO, "Attributes to match: " + this.attmatch.size() + ""); |
---|
251 | |
---|
252 | Iterator it = this.attributes.entrySet().iterator(); |
---|
253 | int j = 0; |
---|
254 | while (it.hasNext()) { |
---|
255 | Map.Entry values = (Map.Entry)it.next(); |
---|
256 | ni.setValue(testdata.attribute(j), test.value((int)values.getValue())); |
---|
257 | j++; |
---|
258 | |
---|
259 | } |
---|
260 | |
---|
261 | ni.setClassValue(test.value(test.classAttribute())); |
---|
262 | |
---|
263 | //System.out.println(ni); |
---|
264 | return ni; |
---|
265 | } |
---|
266 | |
---|
267 | /** |
---|
268 | * returns a new instances array with the metric matched training data |
---|
269 | * |
---|
270 | * @return instances |
---|
271 | */ |
---|
272 | public Instances getMatchedTrain() { |
---|
273 | return this.getMatchedInstances("train", this.train); |
---|
274 | } |
---|
275 | |
---|
276 | /** |
---|
277 | * returns a new instances array with the metric matched test data |
---|
278 | * |
---|
279 | * @return instances |
---|
280 | */ |
---|
281 | public Instances getMatchedTest() { |
---|
282 | return this.getMatchedInstances("test", this.test); |
---|
283 | } |
---|
284 | |
---|
285 | // https://weka.wikispaces.com/Programmatic+Use |
---|
286 | private Instances getMatchedInstances(String name, Instances data) { |
---|
287 | // construct our new attributes |
---|
288 | Attribute[] attrs = new Attribute[this.attributes.size()+1]; |
---|
289 | FastVector fwTrain = new FastVector(this.attributes.size()); |
---|
290 | for (int i=0; i < this.attributes.size(); i++) { |
---|
291 | attrs[i] = new Attribute(String.valueOf(i)); |
---|
292 | fwTrain.addElement(attrs[i]); |
---|
293 | } |
---|
294 | // add our ClassAttribute (which is not numeric!) |
---|
295 | ArrayList<String> acl= new ArrayList<String>(); |
---|
296 | acl.add("0"); |
---|
297 | acl.add("1"); |
---|
298 | |
---|
299 | fwTrain.addElement(new Attribute("bug", acl)); |
---|
300 | Instances newTrain = new Instances(name, fwTrain, data.size()); |
---|
301 | newTrain.setClassIndex(newTrain.numAttributes()-1); |
---|
302 | |
---|
303 | for (int i=0; i < data.size(); i++) { |
---|
304 | Instance ni = new DenseInstance(this.attributes.size()+1); |
---|
305 | |
---|
306 | Iterator it = this.attributes.entrySet().iterator(); |
---|
307 | int j = 0; |
---|
308 | while (it.hasNext()) { |
---|
309 | Map.Entry values = (Map.Entry)it.next(); |
---|
310 | int value = (int)values.getValue(); |
---|
311 | |
---|
312 | // key ist traindata |
---|
313 | if (name.equals("train")) { |
---|
314 | value = (int)values.getKey(); |
---|
315 | } |
---|
316 | ni.setValue(newTrain.attribute(j), data.instance(i).value(value)); |
---|
317 | j++; |
---|
318 | } |
---|
319 | ni.setValue(ni.numAttributes()-1, data.instance(i).value(data.classAttribute())); |
---|
320 | |
---|
321 | newTrain.add(ni); |
---|
322 | } |
---|
323 | |
---|
324 | return newTrain; |
---|
325 | } |
---|
326 | |
---|
327 | |
---|
328 | /** |
---|
329 | * performs the attribute selection |
---|
330 | * we perform attribute significance tests and drop attributes |
---|
331 | */ |
---|
332 | public void attributeSelection() throws Exception { |
---|
333 | //Console.traceln(Level.INFO, "Attribute Selection on Training Attributes ("+this.train.numAttributes()+")"); |
---|
334 | this.attributeSelection(this.train); |
---|
335 | //Console.traceln(Level.INFO, "-----"); |
---|
336 | //Console.traceln(Level.INFO, "Attribute Selection on Test Attributes ("+this.test.numAttributes()+")"); |
---|
337 | this.attributeSelection(this.test); |
---|
338 | //Console.traceln(Level.INFO, "-----"); |
---|
339 | } |
---|
340 | |
---|
341 | private void attributeSelection(Instances which) throws Exception { |
---|
342 | // 1. step we have to categorize the attributes |
---|
343 | //http://weka.sourceforge.net/doc.packages/probabilisticSignificanceAE/weka/attributeSelection/SignificanceAttributeEval.html |
---|
344 | |
---|
345 | SignificanceAttributeEval et = new SignificanceAttributeEval(); |
---|
346 | et.buildEvaluator(which); |
---|
347 | //double tmp[] = new double[this.train.numAttributes()]; |
---|
348 | HashMap<String,Double> saeval = new HashMap<String,Double>(); |
---|
349 | // evaluate all training attributes |
---|
350 | // select top 15% of metrics |
---|
351 | for(int i=0; i < which.numAttributes(); i++) { |
---|
352 | if(which.classIndex() != i) { |
---|
353 | saeval.put(which.attribute(i).name(), et.evaluateAttribute(i)); |
---|
354 | } |
---|
355 | //Console.traceln(Level.SEVERE, "Significance Attribute Eval: " + tmp); |
---|
356 | } |
---|
357 | |
---|
358 | HashMap<String, Double> sorted = sortByValues(saeval); |
---|
359 | |
---|
360 | // die letzen 15% wollen wir haben |
---|
361 | float last = ((float)saeval.size() / 100) * 15; |
---|
362 | int drop_first = saeval.size() - (int)last; |
---|
363 | |
---|
364 | //Console.traceln(Level.INFO, "Dropping "+ drop_first + " of " + sorted.size() + " attributes (we only keep the best 15% "+last+")"); |
---|
365 | /* |
---|
366 | Iterator it = sorted.entrySet().iterator(); |
---|
367 | while (it.hasNext()) { |
---|
368 | Map.Entry pair = (Map.Entry)it.next(); |
---|
369 | Console.traceln(Level.INFO, "key: "+(int)pair.getKey()+", value: " + (double)pair.getValue() + ""); |
---|
370 | }*/ |
---|
371 | |
---|
372 | // drop attributes above last |
---|
373 | Iterator it = sorted.entrySet().iterator(); |
---|
374 | while (drop_first > 0) { |
---|
375 | Map.Entry pair = (Map.Entry)it.next(); |
---|
376 | if(which.attribute((String)pair.getKey()).index() != which.classIndex()) { |
---|
377 | |
---|
378 | which.deleteAttributeAt(which.attribute((String)pair.getKey()).index()); |
---|
379 | //Console.traceln(Level.INFO, "dropping attribute: "+ (String)pair.getKey()); |
---|
380 | } |
---|
381 | drop_first-=1; |
---|
382 | |
---|
383 | |
---|
384 | } |
---|
385 | // //Console.traceln(Level.INFO, "Now we have " + which.numAttributes() + " attributes left (incl. class attribute!)"); |
---|
386 | } |
---|
387 | |
---|
388 | |
---|
389 | private HashMap sortByValues(HashMap map) { |
---|
390 | List list = new LinkedList(map.entrySet()); |
---|
391 | |
---|
392 | Collections.sort(list, new Comparator() { |
---|
393 | public int compare(Object o1, Object o2) { |
---|
394 | return ((Comparable) ((Map.Entry) (o1)).getValue()) |
---|
395 | .compareTo(((Map.Entry) (o2)).getValue()); |
---|
396 | } |
---|
397 | }); |
---|
398 | |
---|
399 | |
---|
400 | HashMap sortedHashMap = new LinkedHashMap(); |
---|
401 | for (Iterator it = list.iterator(); it.hasNext();) { |
---|
402 | Map.Entry entry = (Map.Entry) it.next(); |
---|
403 | sortedHashMap.put(entry.getKey(), entry.getValue()); |
---|
404 | } |
---|
405 | return sortedHashMap; |
---|
406 | } |
---|
407 | |
---|
408 | /** |
---|
409 | * Calculates the Percentiles of the source and target metrics. |
---|
410 | * |
---|
411 | * @param cutoff |
---|
412 | */ |
---|
413 | public void percentiles(double cutoff) { |
---|
414 | for( int i = 0; i < this.train.numAttributes()-1; i++ ) { |
---|
415 | for( int j = 0; j < this.test.numAttributes()-1; j++ ) { |
---|
416 | // class attributes are not relevant |
---|
417 | if( this.train.classIndex() == i ) { |
---|
418 | continue; |
---|
419 | } |
---|
420 | if( this.test.classIndex() == j ) { |
---|
421 | continue; |
---|
422 | } |
---|
423 | |
---|
424 | |
---|
425 | if( !this.attributes.containsKey(i) ) { |
---|
426 | // get percentiles |
---|
427 | double train[] = this.train_values.get(i); |
---|
428 | double test[] = this.test_values.get(j); |
---|
429 | |
---|
430 | Arrays.sort(train); |
---|
431 | Arrays.sort(test); |
---|
432 | |
---|
433 | // percentiles |
---|
434 | double train_p; |
---|
435 | double test_p; |
---|
436 | double score = 0.0; |
---|
437 | for( double p=0.1; p < 1; p+=0.1 ) { |
---|
438 | train_p = train[(int)Math.ceil(train.length * p)]; |
---|
439 | test_p = test[(int)Math.ceil(test.length * p)]; |
---|
440 | |
---|
441 | if( train_p > test_p ) { |
---|
442 | score += test_p / train_p; |
---|
443 | }else { |
---|
444 | score += train_p / test_p; |
---|
445 | } |
---|
446 | } |
---|
447 | |
---|
448 | if( score > cutoff ) { |
---|
449 | this.attributes.put(i, j); |
---|
450 | } |
---|
451 | } |
---|
452 | } |
---|
453 | } |
---|
454 | } |
---|
455 | |
---|
456 | /** |
---|
457 | * calculate Spearmans rank correlation coefficient as matching score |
---|
458 | * |
---|
459 | * @param cutoff |
---|
460 | */ |
---|
461 | public void spearmansRankCorrelation(double cutoff) { |
---|
462 | double p = 0; |
---|
463 | SpearmansCorrelation t = new SpearmansCorrelation(); |
---|
464 | |
---|
465 | // size has to be the same so we randomly sample the number of the smaller sample from the big sample |
---|
466 | if (this.train.size() > this.test.size()) { |
---|
467 | this.sample(this.train, this.test, this.train_values); |
---|
468 | }else if (this.test.size() > this.train.size()) { |
---|
469 | this.sample(this.test, this.train, this.test_values); |
---|
470 | } |
---|
471 | |
---|
472 | // try out possible attribute combinations |
---|
473 | for (int i=0; i < this.train.numAttributes()-1; i++) { |
---|
474 | for (int j=0; j < this.test.numAttributes()-1; j++) { |
---|
475 | // class attributes are not relevant |
---|
476 | if (this.train.classIndex() == i) { |
---|
477 | continue; |
---|
478 | } |
---|
479 | if (this.test.classIndex() == j) { |
---|
480 | continue; |
---|
481 | } |
---|
482 | |
---|
483 | |
---|
484 | if (!this.attributes.containsKey(i)) { |
---|
485 | p = t.correlation(this.train_values.get(i), this.test_values.get(j)); |
---|
486 | if (p > cutoff) { |
---|
487 | this.attributes.put(i, j); |
---|
488 | } |
---|
489 | } |
---|
490 | } |
---|
491 | } |
---|
492 | } |
---|
493 | |
---|
494 | |
---|
495 | public void sample(Instances bigger, Instances smaller, ArrayList<double[]> values) { |
---|
496 | // we want to at keep the indices we select the same |
---|
497 | int indices_to_draw = smaller.size(); |
---|
498 | ArrayList<Integer> indices = new ArrayList<Integer>(); |
---|
499 | Random rand = new Random(); |
---|
500 | while (indices_to_draw > 0) { |
---|
501 | |
---|
502 | int index = rand.nextInt(bigger.size()-1); |
---|
503 | |
---|
504 | if (!indices.contains(index)) { |
---|
505 | indices.add(index); |
---|
506 | indices_to_draw--; |
---|
507 | } |
---|
508 | } |
---|
509 | |
---|
510 | // now reduce our values to the indices we choose above for every attribute |
---|
511 | for (int att=0; att < bigger.numAttributes()-1; att++) { |
---|
512 | |
---|
513 | // get double for the att |
---|
514 | double[] vals = values.get(att); |
---|
515 | double[] new_vals = new double[indices.size()]; |
---|
516 | |
---|
517 | int i = 0; |
---|
518 | for (Iterator<Integer> it = indices.iterator(); it.hasNext();) { |
---|
519 | new_vals[i] = vals[it.next()]; |
---|
520 | i++; |
---|
521 | } |
---|
522 | |
---|
523 | values.set(att, new_vals); |
---|
524 | } |
---|
525 | } |
---|
526 | |
---|
527 | |
---|
528 | /** |
---|
529 | * We run the kolmogorov-smirnov test on the data from our test an traindata |
---|
530 | * if the p value is above the cutoff we include it in the results |
---|
531 | * p value tends to be 0 when the distributions of the data are significantly different |
---|
532 | * but we want them to be the same |
---|
533 | * |
---|
534 | * @param cutoff |
---|
535 | * @return p-val |
---|
536 | */ |
---|
537 | public void kolmogorovSmirnovTest(double cutoff) { |
---|
538 | double p = 0; |
---|
539 | |
---|
540 | KolmogorovSmirnovTest t = new KolmogorovSmirnovTest(); |
---|
541 | |
---|
542 | // todo: this should be symmetrical we don't have to compare i to j and then j to i |
---|
543 | // todo: this relies on the last attribute being the class, |
---|
544 | //Console.traceln(Level.INFO, "Starting Kolmogorov-Smirnov test for traindata size: " + this.train.size() + " attributes("+this.train.numAttributes()+") and testdata size: " + this.test.size() + " attributes("+this.test.numAttributes()+")"); |
---|
545 | for (int i=0; i < this.train.numAttributes()-1; i++) { |
---|
546 | for ( int j=0; j < this.test.numAttributes()-1; j++) { |
---|
547 | //p = t.kolmogorovSmirnovTest(this.train_values.get(i), this.test_values.get(j)); |
---|
548 | //p = t.kolmogorovSmirnovTest(this.train_values.get(i), this.test_values.get(j)); |
---|
549 | // class attributes are not relevant |
---|
550 | if ( this.train.classIndex() == i ) { |
---|
551 | continue; |
---|
552 | } |
---|
553 | if ( this.test.classIndex() == j ) { |
---|
554 | continue; |
---|
555 | } |
---|
556 | // PRoblem: exactP is forced for small sample sizes and it never finishes |
---|
557 | if (!this.attributes.containsKey(i)) { |
---|
558 | |
---|
559 | // todo: output the values and complain on the math.commons mailinglist |
---|
560 | p = t.kolmogorovSmirnovTest(this.train_values.get(i), this.test_values.get(j)); |
---|
561 | if (p > cutoff) { |
---|
562 | this.attributes.put(i, j); |
---|
563 | } |
---|
564 | } |
---|
565 | } |
---|
566 | } |
---|
567 | |
---|
568 | //Console.traceln(Level.INFO, "Found " + this.attmatch.size() + " matching attributes"); |
---|
569 | } |
---|
570 | } |
---|
571 | } |
---|