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 | * SimpleLinearRegression.java |
---|
19 | * Copyright (C) 2002 University of Waikato, Hamilton, New Zealand |
---|
20 | * |
---|
21 | */ |
---|
22 | |
---|
23 | package weka.classifiers.functions; |
---|
24 | |
---|
25 | import weka.classifiers.Classifier; |
---|
26 | import weka.classifiers.AbstractClassifier; |
---|
27 | import weka.core.Attribute; |
---|
28 | import weka.core.Capabilities; |
---|
29 | import weka.core.Instance; |
---|
30 | import weka.core.Instances; |
---|
31 | import weka.core.RevisionUtils; |
---|
32 | import weka.core.Utils; |
---|
33 | import weka.core.WeightedInstancesHandler; |
---|
34 | import weka.core.Capabilities.Capability; |
---|
35 | |
---|
36 | /** |
---|
37 | <!-- globalinfo-start --> |
---|
38 | * Learns a simple linear regression model. Picks the attribute that results in the lowest squared error. Missing values are not allowed. Can only deal with numeric attributes. |
---|
39 | * <p/> |
---|
40 | <!-- globalinfo-end --> |
---|
41 | * |
---|
42 | <!-- options-start --> |
---|
43 | * Valid options are: <p/> |
---|
44 | * |
---|
45 | * <pre> -D |
---|
46 | * If set, classifier is run in debug mode and |
---|
47 | * may output additional info to the console</pre> |
---|
48 | * |
---|
49 | <!-- options-end --> |
---|
50 | * |
---|
51 | * @author Eibe Frank (eibe@cs.waikato.ac.nz) |
---|
52 | * @version $Revision: 5928 $ |
---|
53 | */ |
---|
54 | public class SimpleLinearRegression extends AbstractClassifier |
---|
55 | implements WeightedInstancesHandler { |
---|
56 | |
---|
57 | /** for serialization */ |
---|
58 | static final long serialVersionUID = 1679336022895414137L; |
---|
59 | |
---|
60 | /** The chosen attribute */ |
---|
61 | private Attribute m_attribute; |
---|
62 | |
---|
63 | /** The index of the chosen attribute */ |
---|
64 | private int m_attributeIndex; |
---|
65 | |
---|
66 | /** The slope */ |
---|
67 | private double m_slope; |
---|
68 | |
---|
69 | /** The intercept */ |
---|
70 | private double m_intercept; |
---|
71 | |
---|
72 | /** If true, suppress error message if no useful attribute was found*/ |
---|
73 | private boolean m_suppressErrorMessage = false; |
---|
74 | |
---|
75 | /** |
---|
76 | * Returns a string describing this classifier |
---|
77 | * @return a description of the classifier suitable for |
---|
78 | * displaying in the explorer/experimenter gui |
---|
79 | */ |
---|
80 | public String globalInfo() { |
---|
81 | return "Learns a simple linear regression model. " |
---|
82 | +"Picks the attribute that results in the lowest squared error. " |
---|
83 | +"Missing values are not allowed. Can only deal with numeric attributes."; |
---|
84 | } |
---|
85 | |
---|
86 | /** |
---|
87 | * Generate a prediction for the supplied instance. |
---|
88 | * |
---|
89 | * @param inst the instance to predict. |
---|
90 | * @return the prediction |
---|
91 | * @throws Exception if an error occurs |
---|
92 | */ |
---|
93 | public double classifyInstance(Instance inst) throws Exception { |
---|
94 | |
---|
95 | if (m_attribute == null) { |
---|
96 | return m_intercept; |
---|
97 | } else { |
---|
98 | if (inst.isMissing(m_attribute.index())) { |
---|
99 | throw new Exception("SimpleLinearRegression: No missing values!"); |
---|
100 | } |
---|
101 | return m_intercept + m_slope * inst.value(m_attribute.index()); |
---|
102 | } |
---|
103 | } |
---|
104 | |
---|
105 | /** |
---|
106 | * Returns default capabilities of the classifier. |
---|
107 | * |
---|
108 | * @return the capabilities of this classifier |
---|
109 | */ |
---|
110 | public Capabilities getCapabilities() { |
---|
111 | Capabilities result = super.getCapabilities(); |
---|
112 | result.disableAll(); |
---|
113 | |
---|
114 | // attributes |
---|
115 | result.enable(Capability.NUMERIC_ATTRIBUTES); |
---|
116 | result.enable(Capability.DATE_ATTRIBUTES); |
---|
117 | |
---|
118 | // class |
---|
119 | result.enable(Capability.NUMERIC_CLASS); |
---|
120 | result.enable(Capability.DATE_CLASS); |
---|
121 | result.enable(Capability.MISSING_CLASS_VALUES); |
---|
122 | |
---|
123 | return result; |
---|
124 | } |
---|
125 | |
---|
126 | /** |
---|
127 | * Builds a simple linear regression model given the supplied training data. |
---|
128 | * |
---|
129 | * @param insts the training data. |
---|
130 | * @throws Exception if an error occurs |
---|
131 | */ |
---|
132 | public void buildClassifier(Instances insts) throws Exception { |
---|
133 | |
---|
134 | // can classifier handle the data? |
---|
135 | getCapabilities().testWithFail(insts); |
---|
136 | |
---|
137 | // remove instances with missing class |
---|
138 | insts = new Instances(insts); |
---|
139 | insts.deleteWithMissingClass(); |
---|
140 | |
---|
141 | // Compute mean of target value |
---|
142 | double yMean = insts.meanOrMode(insts.classIndex()); |
---|
143 | |
---|
144 | // Choose best attribute |
---|
145 | double minMsq = Double.MAX_VALUE; |
---|
146 | m_attribute = null; |
---|
147 | int chosen = -1; |
---|
148 | double chosenSlope = Double.NaN; |
---|
149 | double chosenIntercept = Double.NaN; |
---|
150 | for (int i = 0; i < insts.numAttributes(); i++) { |
---|
151 | if (i != insts.classIndex()) { |
---|
152 | m_attribute = insts.attribute(i); |
---|
153 | |
---|
154 | // Compute slope and intercept |
---|
155 | double xMean = insts.meanOrMode(i); |
---|
156 | double sumWeightedXDiffSquared = 0; |
---|
157 | double sumWeightedYDiffSquared = 0; |
---|
158 | m_slope = 0; |
---|
159 | for (int j = 0; j < insts.numInstances(); j++) { |
---|
160 | Instance inst = insts.instance(j); |
---|
161 | if (!inst.isMissing(i) && !inst.classIsMissing()) { |
---|
162 | double xDiff = inst.value(i) - xMean; |
---|
163 | double yDiff = inst.classValue() - yMean; |
---|
164 | double weightedXDiff = inst.weight() * xDiff; |
---|
165 | double weightedYDiff = inst.weight() * yDiff; |
---|
166 | m_slope += weightedXDiff * yDiff; |
---|
167 | sumWeightedXDiffSquared += weightedXDiff * xDiff; |
---|
168 | sumWeightedYDiffSquared += weightedYDiff * yDiff; |
---|
169 | } |
---|
170 | } |
---|
171 | |
---|
172 | // Skip attribute if not useful |
---|
173 | if (sumWeightedXDiffSquared == 0) { |
---|
174 | continue; |
---|
175 | } |
---|
176 | double numerator = m_slope; |
---|
177 | m_slope /= sumWeightedXDiffSquared; |
---|
178 | m_intercept = yMean - m_slope * xMean; |
---|
179 | |
---|
180 | // Compute sum of squared errors |
---|
181 | double msq = sumWeightedYDiffSquared - m_slope * numerator; |
---|
182 | |
---|
183 | // Check whether this is the best attribute |
---|
184 | if (msq < minMsq) { |
---|
185 | minMsq = msq; |
---|
186 | chosen = i; |
---|
187 | chosenSlope = m_slope; |
---|
188 | chosenIntercept = m_intercept; |
---|
189 | } |
---|
190 | } |
---|
191 | } |
---|
192 | |
---|
193 | // Set parameters |
---|
194 | if (chosen == -1) { |
---|
195 | if (!m_suppressErrorMessage) System.err.println("----- no useful attribute found"); |
---|
196 | m_attribute = null; |
---|
197 | m_attributeIndex = 0; |
---|
198 | m_slope = 0; |
---|
199 | m_intercept = yMean; |
---|
200 | } else { |
---|
201 | m_attribute = insts.attribute(chosen); |
---|
202 | m_attributeIndex = chosen; |
---|
203 | m_slope = chosenSlope; |
---|
204 | m_intercept = chosenIntercept; |
---|
205 | } |
---|
206 | } |
---|
207 | |
---|
208 | /** |
---|
209 | * Returns true if a usable attribute was found. |
---|
210 | * |
---|
211 | * @return true if a usable attribute was found. |
---|
212 | */ |
---|
213 | public boolean foundUsefulAttribute(){ |
---|
214 | return (m_attribute != null); |
---|
215 | } |
---|
216 | |
---|
217 | /** |
---|
218 | * Returns the index of the attribute used in the regression. |
---|
219 | * |
---|
220 | * @return the index of the attribute. |
---|
221 | */ |
---|
222 | public int getAttributeIndex(){ |
---|
223 | return m_attributeIndex; |
---|
224 | } |
---|
225 | |
---|
226 | /** |
---|
227 | * Returns the slope of the function. |
---|
228 | * |
---|
229 | * @return the slope. |
---|
230 | */ |
---|
231 | public double getSlope(){ |
---|
232 | return m_slope; |
---|
233 | } |
---|
234 | |
---|
235 | /** |
---|
236 | * Returns the intercept of the function. |
---|
237 | * |
---|
238 | * @return the intercept. |
---|
239 | */ |
---|
240 | public double getIntercept(){ |
---|
241 | return m_intercept; |
---|
242 | } |
---|
243 | |
---|
244 | /** |
---|
245 | * Turn off the error message that is reported when no useful attribute is found. |
---|
246 | * |
---|
247 | * @param s if set to true turns off the error message |
---|
248 | */ |
---|
249 | public void setSuppressErrorMessage(boolean s){ |
---|
250 | m_suppressErrorMessage = s; |
---|
251 | } |
---|
252 | |
---|
253 | /** |
---|
254 | * Returns a description of this classifier as a string |
---|
255 | * |
---|
256 | * @return a description of the classifier. |
---|
257 | */ |
---|
258 | public String toString() { |
---|
259 | |
---|
260 | StringBuffer text = new StringBuffer(); |
---|
261 | if (m_attribute == null) { |
---|
262 | text.append("Predicting constant " + m_intercept); |
---|
263 | } else { |
---|
264 | text.append("Linear regression on " + m_attribute.name() + "\n\n"); |
---|
265 | text.append(Utils.doubleToString(m_slope,2) + " * " + |
---|
266 | m_attribute.name()); |
---|
267 | if (m_intercept > 0) { |
---|
268 | text.append(" + " + Utils.doubleToString(m_intercept, 2)); |
---|
269 | } else { |
---|
270 | text.append(" - " + Utils.doubleToString((-m_intercept), 2)); |
---|
271 | } |
---|
272 | } |
---|
273 | text.append("\n"); |
---|
274 | return text.toString(); |
---|
275 | } |
---|
276 | |
---|
277 | /** |
---|
278 | * Returns the revision string. |
---|
279 | * |
---|
280 | * @return the revision |
---|
281 | */ |
---|
282 | public String getRevision() { |
---|
283 | return RevisionUtils.extract("$Revision: 5928 $"); |
---|
284 | } |
---|
285 | |
---|
286 | /** |
---|
287 | * Main method for testing this class |
---|
288 | * |
---|
289 | * @param argv options |
---|
290 | */ |
---|
291 | public static void main(String [] argv){ |
---|
292 | runClassifier(new SimpleLinearRegression(), argv); |
---|
293 | } |
---|
294 | } |
---|