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 | * FilteredClassifier.java
|
---|
19 | * Copyright (C) 1999 Len Trigg
|
---|
20 | *
|
---|
21 | */
|
---|
22 |
|
---|
23 | package weka.classifiers;
|
---|
24 |
|
---|
25 | import java.util.Enumeration;
|
---|
26 | import java.util.Vector;
|
---|
27 | import weka.core.Instance;
|
---|
28 | import weka.core.Instances;
|
---|
29 | import weka.core.Option;
|
---|
30 | import weka.core.OptionHandler;
|
---|
31 | import weka.core.Utils;
|
---|
32 | import weka.filters.Filter;
|
---|
33 | import weka.core.Attribute;
|
---|
34 |
|
---|
35 |
|
---|
36 | /**
|
---|
37 | * Class for running an arbitrary classifier on data that has been passed
|
---|
38 | * through an arbitrary filter.<p>
|
---|
39 | *
|
---|
40 | * Valid options from the command line are:<p>
|
---|
41 | *
|
---|
42 | * -B classifierstring <br>
|
---|
43 | * Classifierstring should contain the full class name of a classifier
|
---|
44 | * followed by options to the classifier.
|
---|
45 | * (required).<p>
|
---|
46 | *
|
---|
47 | * -F filterstring <br>
|
---|
48 | * Filterstring should contain the full class name of a filter
|
---|
49 | * followed by options to the filter.
|
---|
50 | * (required).<p>
|
---|
51 | *
|
---|
52 | * @author Len Trigg ([email protected])
|
---|
53 | * @version $Revision: 8815 $
|
---|
54 | */
|
---|
55 | public class FilteredClassifier extends DistributionClassifier
|
---|
56 | implements OptionHandler {
|
---|
57 |
|
---|
58 | /** The classifier */
|
---|
59 | protected Classifier m_Classifier = new weka.classifiers.NaiveBayesSimple();
|
---|
60 |
|
---|
61 | /** The filter */
|
---|
62 | protected Filter m_Filter = new weka.filters.DiscretizeFilter();
|
---|
63 |
|
---|
64 | /** The instance structure of the filtered instances */
|
---|
65 | protected Instances m_FilteredInstances;
|
---|
66 |
|
---|
67 | /**
|
---|
68 | * Default constructor specifying NaiveBayesSimple as the classifier and
|
---|
69 | * DiscretizeFilter as the filter. Both of these are just placeholders
|
---|
70 | * for more useful selections.
|
---|
71 | */
|
---|
72 | public FilteredClassifier() {
|
---|
73 |
|
---|
74 | this(new weka.classifiers.NaiveBayesSimple(),
|
---|
75 | new weka.filters.DiscretizeFilter());
|
---|
76 | }
|
---|
77 |
|
---|
78 | /**
|
---|
79 | * Constructor that specifies the subclassifier and filter to use.
|
---|
80 | *
|
---|
81 | * @param classifier the Classifier to receive filtered instances.
|
---|
82 | * @param filter the Filter that will process instances before
|
---|
83 | * passing to the Classifier.
|
---|
84 | */
|
---|
85 | public FilteredClassifier(Classifier classifier, Filter filter) {
|
---|
86 |
|
---|
87 | m_Classifier = classifier;
|
---|
88 | m_Filter = filter;
|
---|
89 | }
|
---|
90 |
|
---|
91 | /**
|
---|
92 | * Returns an enumeration describing the available options
|
---|
93 | *
|
---|
94 | * @return an enumeration of all the available options
|
---|
95 | */
|
---|
96 | public Enumeration listOptions() {
|
---|
97 |
|
---|
98 | Vector newVector = new Vector(2);
|
---|
99 |
|
---|
100 | newVector.addElement(new Option(
|
---|
101 | "\tFull class name of classifier to use, followed\n"
|
---|
102 | + "\tby scheme options. (required)\n"
|
---|
103 | + "\teg: \"weka.classifiers.NaiveBayes -D\"",
|
---|
104 | "B", 1, "-B <classifier specification>"));
|
---|
105 | newVector.addElement(new Option(
|
---|
106 | "\tFull class name of filter to use, followed\n"
|
---|
107 | + "\tby filter options. (required)\n"
|
---|
108 | + "\teg: \"weka.filters.AttributeFilter -V -R 1,2\"",
|
---|
109 | "F", 1, "-F <filter specification>"));
|
---|
110 | return newVector.elements();
|
---|
111 | }
|
---|
112 |
|
---|
113 | /**
|
---|
114 | * Parses a given list of options. Valid options are:<p>
|
---|
115 | *
|
---|
116 | * -B classifierstring <br>
|
---|
117 | * Classifierstring should contain the full class name of a classifier
|
---|
118 | * followed by options to the classifier.
|
---|
119 | * (required).<p>
|
---|
120 | *
|
---|
121 | * -F filterstring <br>
|
---|
122 | * Filterstring should contain the full class name of a filter
|
---|
123 | * followed by options to the filter.
|
---|
124 | * (required).<p>
|
---|
125 | *
|
---|
126 | * @param options the list of options as an array of strings
|
---|
127 | * @exception Exception if an option is not supported
|
---|
128 | */
|
---|
129 | public void setOptions(String[] options) throws Exception {
|
---|
130 |
|
---|
131 | String classifierString = Utils.getOption('B', options);
|
---|
132 | if (classifierString.length() == 0) {
|
---|
133 | throw new Exception("A classifier must be specified"
|
---|
134 | + " with the -B option.");
|
---|
135 | }
|
---|
136 | String [] classifierSpec = Utils.splitOptions(classifierString);
|
---|
137 | if (classifierSpec.length == 0) {
|
---|
138 | throw new Exception("Invalid classifier specification string");
|
---|
139 | }
|
---|
140 | String classifierName = classifierSpec[0];
|
---|
141 | classifierSpec[0] = "";
|
---|
142 | setClassifier(Classifier.forName(classifierName, classifierSpec));
|
---|
143 |
|
---|
144 | // Same for filter
|
---|
145 | String filterString = Utils.getOption('F', options);
|
---|
146 | if (filterString.length() == 0) {
|
---|
147 | throw new Exception("A filter must be specified"
|
---|
148 | + " with the -F option.");
|
---|
149 | }
|
---|
150 | String [] filterSpec = Utils.splitOptions(filterString);
|
---|
151 | if (filterSpec.length == 0) {
|
---|
152 | throw new Exception("Invalid filter specification string");
|
---|
153 | }
|
---|
154 | String filterName = filterSpec[0];
|
---|
155 | filterSpec[0] = "";
|
---|
156 | setFilter((Filter) Utils.forName(Filter.class, filterName, filterSpec));
|
---|
157 | }
|
---|
158 |
|
---|
159 | /**
|
---|
160 | * Gets the current settings of the Classifier.
|
---|
161 | *
|
---|
162 | * @return an array of strings suitable for passing to setOptions
|
---|
163 | */
|
---|
164 | public String [] getOptions() {
|
---|
165 |
|
---|
166 | String [] options = new String [4];
|
---|
167 | int current = 0;
|
---|
168 |
|
---|
169 | options[current++] = "-B";
|
---|
170 | options[current++] = "" + getClassifierSpec();
|
---|
171 |
|
---|
172 | // Same for filter
|
---|
173 | options[current++] = "-F";
|
---|
174 | options[current++] = "" + getFilterSpec();
|
---|
175 |
|
---|
176 | while (current < options.length) {
|
---|
177 | options[current++] = "";
|
---|
178 | }
|
---|
179 | return options;
|
---|
180 | }
|
---|
181 |
|
---|
182 | /**
|
---|
183 | * Sets the classifier
|
---|
184 | *
|
---|
185 | * @param classifier the classifier with all options set.
|
---|
186 | */
|
---|
187 | public void setClassifier(Classifier classifier) {
|
---|
188 |
|
---|
189 | m_Classifier = classifier;
|
---|
190 | }
|
---|
191 |
|
---|
192 | /**
|
---|
193 | * Gets the classifier used.
|
---|
194 | *
|
---|
195 | * @return the classifier
|
---|
196 | */
|
---|
197 | public Classifier getClassifier() {
|
---|
198 |
|
---|
199 | return m_Classifier;
|
---|
200 | }
|
---|
201 |
|
---|
202 | /**
|
---|
203 | * Gets the classifier specification string, which contains the class name of
|
---|
204 | * the classifier and any options to the classifier
|
---|
205 | *
|
---|
206 | * @return the classifier string.
|
---|
207 | */
|
---|
208 | protected String getClassifierSpec() {
|
---|
209 |
|
---|
210 | Classifier c = getClassifier();
|
---|
211 | if (c instanceof OptionHandler) {
|
---|
212 | return c.getClass().getName() + " "
|
---|
213 | + Utils.joinOptions(((OptionHandler)c).getOptions());
|
---|
214 | }
|
---|
215 | return c.getClass().getName();
|
---|
216 | }
|
---|
217 |
|
---|
218 | /**
|
---|
219 | * Sets the filter
|
---|
220 | *
|
---|
221 | * @param filter the filter with all options set.
|
---|
222 | */
|
---|
223 | public void setFilter(Filter filter) {
|
---|
224 |
|
---|
225 | m_Filter = filter;
|
---|
226 | }
|
---|
227 |
|
---|
228 | /**
|
---|
229 | * Gets the filter used.
|
---|
230 | *
|
---|
231 | * @return the filter
|
---|
232 | */
|
---|
233 | public Filter getFilter() {
|
---|
234 |
|
---|
235 | return m_Filter;
|
---|
236 | }
|
---|
237 |
|
---|
238 | /**
|
---|
239 | * Gets the filter specification string, which contains the class name of
|
---|
240 | * the filter and any options to the filter
|
---|
241 | *
|
---|
242 | * @return the filter string.
|
---|
243 | */
|
---|
244 | protected String getFilterSpec() {
|
---|
245 |
|
---|
246 | Filter c = getFilter();
|
---|
247 | if (c instanceof OptionHandler) {
|
---|
248 | return c.getClass().getName() + " "
|
---|
249 | + Utils.joinOptions(((OptionHandler)c).getOptions());
|
---|
250 | }
|
---|
251 | return c.getClass().getName();
|
---|
252 | }
|
---|
253 |
|
---|
254 | /**
|
---|
255 | * Build the classifier on the filtered data.
|
---|
256 | *
|
---|
257 | * @param data the training data
|
---|
258 | * @exception Exception if the classifier could not be built successfully
|
---|
259 | */
|
---|
260 | public void buildClassifier(Instances data) throws Exception {
|
---|
261 |
|
---|
262 | if (m_Classifier == null) {
|
---|
263 | throw new Exception("No base classifiers have been set!");
|
---|
264 | }
|
---|
265 | /*
|
---|
266 | String fname = m_Filter.getClass().getName();
|
---|
267 | fname = fname.substring(fname.lastIndexOf('.') + 1);
|
---|
268 | util.Timer t = util.Timer.getTimer("FilteredClassifier::" + fname);
|
---|
269 | t.start();
|
---|
270 | */
|
---|
271 | m_Filter.setInputFormat(data);
|
---|
272 | data = Filter.useFilter(data, m_Filter);
|
---|
273 | //t.stop();
|
---|
274 | m_FilteredInstances = data.stringFreeStructure();
|
---|
275 | m_Classifier.buildClassifier(data);
|
---|
276 | }
|
---|
277 |
|
---|
278 | /**
|
---|
279 | * Classifies a given instance after filtering.
|
---|
280 | *
|
---|
281 | * @param instance the instance to be classified
|
---|
282 | * @exception Exception if instance could not be classified
|
---|
283 | * successfully
|
---|
284 | */
|
---|
285 | public double [] distributionForInstance(Instance instance)
|
---|
286 | throws Exception {
|
---|
287 |
|
---|
288 | /*
|
---|
289 | System.err.println("FilteredClassifier:: "
|
---|
290 | + m_Filter.getClass().getName()
|
---|
291 | + " in: " + instance);
|
---|
292 | */
|
---|
293 | if (m_Filter.numPendingOutput() > 0) {
|
---|
294 | throw new Exception("Filter output queue not empty!");
|
---|
295 | }
|
---|
296 | /*
|
---|
297 | String fname = m_Filter.getClass().getName();
|
---|
298 | fname = fname.substring(fname.lastIndexOf('.') + 1);
|
---|
299 | util.Timer t = util.Timer.getTimer("FilteredClassifier::" + fname);
|
---|
300 | t.start();
|
---|
301 | */
|
---|
302 | if (!m_Filter.input(instance)) {
|
---|
303 | throw new Exception("Filter didn't make the test instance"
|
---|
304 | + " immediately available!");
|
---|
305 | }
|
---|
306 | m_Filter.batchFinished();
|
---|
307 | Instance newInstance = m_Filter.output();
|
---|
308 | //t.stop();
|
---|
309 | /*
|
---|
310 | System.err.println("FilteredClassifier:: "
|
---|
311 | + m_Filter.getClass().getName()
|
---|
312 | + " out: " + newInstance);
|
---|
313 | */
|
---|
314 | if (m_Classifier instanceof DistributionClassifier) {
|
---|
315 | return ((DistributionClassifier)m_Classifier)
|
---|
316 | .distributionForInstance(newInstance);
|
---|
317 | }
|
---|
318 | double pred = m_Classifier.classifyInstance(newInstance);
|
---|
319 | double [] result = new double [m_FilteredInstances.numClasses()];
|
---|
320 | if (Instance.isMissingValue(pred)) {
|
---|
321 | return result;
|
---|
322 | }
|
---|
323 | switch (instance.classAttribute().type()) {
|
---|
324 | case Attribute.NOMINAL:
|
---|
325 | result[(int) pred] = 1.0;
|
---|
326 | break;
|
---|
327 | case Attribute.NUMERIC:
|
---|
328 | result[0] = pred;
|
---|
329 | break;
|
---|
330 | default:
|
---|
331 | throw new Exception("Unknown class type");
|
---|
332 | }
|
---|
333 | return result;
|
---|
334 | }
|
---|
335 |
|
---|
336 | /**
|
---|
337 | * Output a representation of this classifier
|
---|
338 | */
|
---|
339 | public String toString() {
|
---|
340 |
|
---|
341 | if (m_FilteredInstances == null) {
|
---|
342 | return "FilteredClassifier: No model built yet.";
|
---|
343 | }
|
---|
344 |
|
---|
345 | String result = "FilteredClassifier using "
|
---|
346 | + getClassifierSpec()
|
---|
347 | + " on data filtered through "
|
---|
348 | + getFilterSpec()
|
---|
349 | + "\n\nFiltered Header\n"
|
---|
350 | + m_FilteredInstances.toString()
|
---|
351 | + "\n\nClassifier Model\n"
|
---|
352 | + m_Classifier.toString();
|
---|
353 | return result;
|
---|
354 | }
|
---|
355 |
|
---|
356 |
|
---|
357 | /**
|
---|
358 | * Main method for testing this class.
|
---|
359 | *
|
---|
360 | * @param argv should contain the following arguments:
|
---|
361 | * -t training file [-T test file] [-c class index]
|
---|
362 | */
|
---|
363 | public static void main(String [] argv) {
|
---|
364 |
|
---|
365 | try {
|
---|
366 | System.out.println("Evaluation disabled!");
|
---|
367 | //System.out.println(Evaluation.evaluateModel(new FilteredClassifier(),
|
---|
368 | // argv));
|
---|
369 | } catch (Exception e) {
|
---|
370 | System.err.println(e.getMessage());
|
---|
371 | }
|
---|
372 | }
|
---|
373 |
|
---|
374 | }
|
---|