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 | * BayesNet.java |
---|
19 | * Copyright (C) 2004 University of Waikato, Hamilton, New Zealand |
---|
20 | * |
---|
21 | */ |
---|
22 | |
---|
23 | package weka.classifiers.bayes.net.estimate; |
---|
24 | |
---|
25 | import weka.classifiers.bayes.BayesNet; |
---|
26 | import weka.core.Instance; |
---|
27 | import weka.core.Instances; |
---|
28 | import weka.core.RevisionUtils; |
---|
29 | import weka.core.Utils; |
---|
30 | import weka.estimators.Estimator; |
---|
31 | |
---|
32 | import java.util.Enumeration; |
---|
33 | |
---|
34 | /** |
---|
35 | <!-- globalinfo-start --> |
---|
36 | * SimpleEstimator is used for estimating the conditional probability tables of a Bayes network once the structure has been learned. Estimates probabilities directly from data. |
---|
37 | * <p/> |
---|
38 | <!-- globalinfo-end --> |
---|
39 | * |
---|
40 | <!-- options-start --> |
---|
41 | * Valid options are: <p/> |
---|
42 | * |
---|
43 | * <pre> -A <alpha> |
---|
44 | * Initial count (alpha) |
---|
45 | * </pre> |
---|
46 | * |
---|
47 | <!-- options-end --> |
---|
48 | * |
---|
49 | * @author Remco Bouckaert (rrb@xm.co.nz) |
---|
50 | * @version $Revision: 1.6 $ |
---|
51 | */ |
---|
52 | public class SimpleEstimator |
---|
53 | extends BayesNetEstimator { |
---|
54 | |
---|
55 | /** for serialization */ |
---|
56 | static final long serialVersionUID = 5874941612331806172L; |
---|
57 | |
---|
58 | /** |
---|
59 | * Returns a string describing this object |
---|
60 | * @return a description of the classifier suitable for |
---|
61 | * displaying in the explorer/experimenter gui |
---|
62 | */ |
---|
63 | public String globalInfo() { |
---|
64 | return |
---|
65 | "SimpleEstimator is used for estimating the conditional probability " |
---|
66 | + "tables of a Bayes network once the structure has been learned. " |
---|
67 | + "Estimates probabilities directly from data."; |
---|
68 | } |
---|
69 | |
---|
70 | /** |
---|
71 | * estimateCPTs estimates the conditional probability tables for the Bayes |
---|
72 | * Net using the network structure. |
---|
73 | * |
---|
74 | * @param bayesNet the bayes net to use |
---|
75 | * @throws Exception if something goes wrong |
---|
76 | */ |
---|
77 | public void estimateCPTs(BayesNet bayesNet) throws Exception { |
---|
78 | initCPTs(bayesNet); |
---|
79 | |
---|
80 | // Compute counts |
---|
81 | Enumeration enumInsts = bayesNet.m_Instances.enumerateInstances(); |
---|
82 | while (enumInsts.hasMoreElements()) { |
---|
83 | Instance instance = (Instance) enumInsts.nextElement(); |
---|
84 | |
---|
85 | updateClassifier(bayesNet, instance); |
---|
86 | } |
---|
87 | } // estimateCPTs |
---|
88 | |
---|
89 | /** |
---|
90 | * Updates the classifier with the given instance. |
---|
91 | * |
---|
92 | * @param bayesNet the bayes net to use |
---|
93 | * @param instance the new training instance to include in the model |
---|
94 | * @throws Exception if the instance could not be incorporated in |
---|
95 | * the model. |
---|
96 | */ |
---|
97 | public void updateClassifier(BayesNet bayesNet, Instance instance) throws Exception { |
---|
98 | for (int iAttribute = 0; iAttribute < bayesNet.m_Instances.numAttributes(); iAttribute++) { |
---|
99 | double iCPT = 0; |
---|
100 | |
---|
101 | for (int iParent = 0; iParent < bayesNet.getParentSet(iAttribute).getNrOfParents(); iParent++) { |
---|
102 | int nParent = bayesNet.getParentSet(iAttribute).getParent(iParent); |
---|
103 | |
---|
104 | iCPT = iCPT * bayesNet.m_Instances.attribute(nParent).numValues() + instance.value(nParent); |
---|
105 | } |
---|
106 | |
---|
107 | bayesNet.m_Distributions[iAttribute][(int) iCPT].addValue(instance.value(iAttribute), instance.weight()); |
---|
108 | } |
---|
109 | } // updateClassifier |
---|
110 | |
---|
111 | |
---|
112 | /** |
---|
113 | * initCPTs reserves space for CPTs and set all counts to zero |
---|
114 | * |
---|
115 | * @param bayesNet the bayes net to use |
---|
116 | * @throws Exception if something goes wrong |
---|
117 | */ |
---|
118 | public void initCPTs(BayesNet bayesNet) throws Exception { |
---|
119 | Instances instances = bayesNet.m_Instances; |
---|
120 | |
---|
121 | // Reserve space for CPTs |
---|
122 | int nMaxParentCardinality = 1; |
---|
123 | for (int iAttribute = 0; iAttribute < instances.numAttributes(); iAttribute++) { |
---|
124 | if (bayesNet.getParentSet(iAttribute).getCardinalityOfParents() > nMaxParentCardinality) { |
---|
125 | nMaxParentCardinality = bayesNet.getParentSet(iAttribute).getCardinalityOfParents(); |
---|
126 | } |
---|
127 | } |
---|
128 | |
---|
129 | // Reserve plenty of memory |
---|
130 | bayesNet.m_Distributions = new Estimator[instances.numAttributes()][nMaxParentCardinality]; |
---|
131 | |
---|
132 | // estimate CPTs |
---|
133 | for (int iAttribute = 0; iAttribute < instances.numAttributes(); iAttribute++) { |
---|
134 | for (int iParent = 0; iParent < bayesNet.getParentSet(iAttribute).getCardinalityOfParents(); iParent++) { |
---|
135 | bayesNet.m_Distributions[iAttribute][iParent] = |
---|
136 | new DiscreteEstimatorBayes(instances.attribute(iAttribute).numValues(), m_fAlpha); |
---|
137 | } |
---|
138 | } |
---|
139 | } // initCPTs |
---|
140 | |
---|
141 | /** |
---|
142 | * Calculates the class membership probabilities for the given test |
---|
143 | * instance. |
---|
144 | * |
---|
145 | * @param bayesNet the bayes net to use |
---|
146 | * @param instance the instance to be classified |
---|
147 | * @return predicted class probability distribution |
---|
148 | * @throws Exception if there is a problem generating the prediction |
---|
149 | */ |
---|
150 | public double[] distributionForInstance(BayesNet bayesNet, Instance instance) throws Exception { |
---|
151 | Instances instances = bayesNet.m_Instances; |
---|
152 | int nNumClasses = instances.numClasses(); |
---|
153 | double[] fProbs = new double[nNumClasses]; |
---|
154 | |
---|
155 | for (int iClass = 0; iClass < nNumClasses; iClass++) { |
---|
156 | fProbs[iClass] = 1.0; |
---|
157 | } |
---|
158 | |
---|
159 | for (int iClass = 0; iClass < nNumClasses; iClass++) { |
---|
160 | double logfP = 0; |
---|
161 | |
---|
162 | for (int iAttribute = 0; iAttribute < instances.numAttributes(); iAttribute++) { |
---|
163 | double iCPT = 0; |
---|
164 | |
---|
165 | for (int iParent = 0; iParent < bayesNet.getParentSet(iAttribute).getNrOfParents(); iParent++) { |
---|
166 | int nParent = bayesNet.getParentSet(iAttribute).getParent(iParent); |
---|
167 | |
---|
168 | if (nParent == instances.classIndex()) { |
---|
169 | iCPT = iCPT * nNumClasses + iClass; |
---|
170 | } else { |
---|
171 | iCPT = iCPT * instances.attribute(nParent).numValues() + instance.value(nParent); |
---|
172 | } |
---|
173 | } |
---|
174 | |
---|
175 | if (iAttribute == instances.classIndex()) { |
---|
176 | // fP *= |
---|
177 | // m_Distributions[iAttribute][(int) iCPT].getProbability(iClass); |
---|
178 | logfP += Math.log(bayesNet.m_Distributions[iAttribute][(int) iCPT].getProbability(iClass)); |
---|
179 | } else { |
---|
180 | // fP *= |
---|
181 | // m_Distributions[iAttribute][(int) iCPT] |
---|
182 | // .getProbability(instance.value(iAttribute)); |
---|
183 | logfP |
---|
184 | += Math.log(bayesNet.m_Distributions[iAttribute][(int) iCPT].getProbability(instance.value(iAttribute))); |
---|
185 | } |
---|
186 | } |
---|
187 | |
---|
188 | // fProbs[iClass] *= fP; |
---|
189 | fProbs[iClass] += logfP; |
---|
190 | } |
---|
191 | |
---|
192 | // Find maximum |
---|
193 | double fMax = fProbs[0]; |
---|
194 | for (int iClass = 0; iClass < nNumClasses; iClass++) { |
---|
195 | if (fProbs[iClass] > fMax) { |
---|
196 | fMax = fProbs[iClass]; |
---|
197 | } |
---|
198 | } |
---|
199 | // transform from log-space to normal-space |
---|
200 | for (int iClass = 0; iClass < nNumClasses; iClass++) { |
---|
201 | fProbs[iClass] = Math.exp(fProbs[iClass] - fMax); |
---|
202 | } |
---|
203 | |
---|
204 | // Display probabilities |
---|
205 | Utils.normalize(fProbs); |
---|
206 | |
---|
207 | return fProbs; |
---|
208 | } // distributionForInstance |
---|
209 | |
---|
210 | /** |
---|
211 | * Returns the revision string. |
---|
212 | * |
---|
213 | * @return the revision |
---|
214 | */ |
---|
215 | public String getRevision() { |
---|
216 | return RevisionUtils.extract("$Revision: 1.6 $"); |
---|
217 | } |
---|
218 | |
---|
219 | } // SimpleEstimator |
---|