@Override
protected pikater.ontology.messages.Evaluation evaluateCA() {
Evaluation eval = test();
pikater.ontology.messages.Evaluation result = new pikater.ontology.messages.Evaluation();
result.setError_rate((float) eval.errorRate());
try {
result.setKappa_statistic((float) eval.kappa());
} catch (Exception e) {
result.setKappa_statistic(-1);
}
result.setMean_absolute_error((float) eval.meanAbsoluteError());
try {
result.setRelative_absolute_error((float) eval.relativeAbsoluteError());
} catch (Exception e) {
result.setRelative_absolute_error(-1);
}
result.setRoot_mean_squared_error((float) eval.rootMeanSquaredError());
result.setRoot_relative_squared_error((float) eval.rootRelativeSquaredError());
return result;
}
/**
* Evaluates a feature subset by cross validation
*
* @param feature_set the subset to be evaluated
* @param num_atts the number of attributes in the subset
* @return the estimated accuracy
* @throws Exception if subset can't be evaluated
*/
protected double estimatePerformance(BitSet feature_set, int num_atts) throws Exception {
m_evaluation = new Evaluation(m_theInstances);
int i;
int[] fs = new int[num_atts];
double[] instA = new double[num_atts];
int classI = m_theInstances.classIndex();
int index = 0;
for (i = 0; i < m_numAttributes; i++) {
if (feature_set.get(i)) {
fs[index++] = i;
}
}
// create new hash table
m_entries = new Hashtable((int) (m_theInstances.numInstances() * 1.5));
// insert instances into the hash table
for (i = 0; i < m_numInstances; i++) {
Instance inst = m_theInstances.instance(i);
for (int j = 0; j < fs.length; j++) {
if (fs[j] == classI) {
instA[j] = Double.MAX_VALUE; // missing for the class
} else if (inst.isMissing(fs[j])) {
instA[j] = Double.MAX_VALUE;
} else {
instA[j] = inst.value(fs[j]);
}
}
insertIntoTable(inst, instA);
}
if (m_CVFolds == 1) {
// calculate leave one out error
for (i = 0; i < m_numInstances; i++) {
Instance inst = m_theInstances.instance(i);
for (int j = 0; j < fs.length; j++) {
if (fs[j] == classI) {
instA[j] = Double.MAX_VALUE; // missing for the class
} else if (inst.isMissing(fs[j])) {
instA[j] = Double.MAX_VALUE;
} else {
instA[j] = inst.value(fs[j]);
}
}
evaluateInstanceLeaveOneOut(inst, instA);
}
} else {
m_theInstances.randomize(m_rr);
m_theInstances.stratify(m_CVFolds);
// calculate 10 fold cross validation error
for (i = 0; i < m_CVFolds; i++) {
Instances insts = m_theInstances.testCV(m_CVFolds, i);
evaluateFoldCV(insts, fs);
}
}
switch (m_evaluationMeasure) {
case EVAL_DEFAULT:
if (m_classIsNominal) {
return m_evaluation.pctCorrect();
}
return -m_evaluation.rootMeanSquaredError();
case EVAL_ACCURACY:
return m_evaluation.pctCorrect();
case EVAL_RMSE:
return -m_evaluation.rootMeanSquaredError();
case EVAL_MAE:
return -m_evaluation.meanAbsoluteError();
case EVAL_AUC:
double[] classPriors = m_evaluation.getClassPriors();
Utils.normalize(classPriors);
double weightedAUC = 0;
for (i = 0; i < m_theInstances.classAttribute().numValues(); i++) {
double tempAUC = m_evaluation.areaUnderROC(i);
if (!Utils.isMissingValue(tempAUC)) {
weightedAUC += (classPriors[i] * tempAUC);
} else {
System.err.println("Undefined AUC!!");
}
}
return weightedAUC;
}
// shouldn't get here
return 0.0;
}
/**
* Gets the results for the supplied train and test datasets. Now performs a deep copy of the
* classifier before it is built and evaluated (just in case the classifier is not initialized
* properly in buildClassifier()).
*
* @param train the training Instances.
* @param test the testing Instances.
* @return the results stored in an array. The objects stored in the array may be Strings,
* Doubles, or null (for the missing value).
* @throws Exception if a problem occurs while getting the results
*/
public Object[] getResult(Instances train, Instances test) throws Exception {
if (train.classAttribute().type() != Attribute.NUMERIC) {
throw new Exception("Class attribute is not numeric!");
}
if (m_Template == null) {
throw new Exception("No classifier has been specified");
}
ThreadMXBean thMonitor = ManagementFactory.getThreadMXBean();
boolean canMeasureCPUTime = thMonitor.isThreadCpuTimeSupported();
if (canMeasureCPUTime && !thMonitor.isThreadCpuTimeEnabled())
thMonitor.setThreadCpuTimeEnabled(true);
int addm = (m_AdditionalMeasures != null) ? m_AdditionalMeasures.length : 0;
Object[] result = new Object[RESULT_SIZE + addm + m_numPluginStatistics];
long thID = Thread.currentThread().getId();
long CPUStartTime = -1,
trainCPUTimeElapsed = -1,
testCPUTimeElapsed = -1,
trainTimeStart,
trainTimeElapsed,
testTimeStart,
testTimeElapsed;
Evaluation eval = new Evaluation(train);
m_Classifier = AbstractClassifier.makeCopy(m_Template);
trainTimeStart = System.currentTimeMillis();
if (canMeasureCPUTime) CPUStartTime = thMonitor.getThreadUserTime(thID);
m_Classifier.buildClassifier(train);
if (canMeasureCPUTime) trainCPUTimeElapsed = thMonitor.getThreadUserTime(thID) - CPUStartTime;
trainTimeElapsed = System.currentTimeMillis() - trainTimeStart;
testTimeStart = System.currentTimeMillis();
if (canMeasureCPUTime) CPUStartTime = thMonitor.getThreadUserTime(thID);
eval.evaluateModel(m_Classifier, test);
if (canMeasureCPUTime) testCPUTimeElapsed = thMonitor.getThreadUserTime(thID) - CPUStartTime;
testTimeElapsed = System.currentTimeMillis() - testTimeStart;
thMonitor = null;
m_result = eval.toSummaryString();
// The results stored are all per instance -- can be multiplied by the
// number of instances to get absolute numbers
int current = 0;
result[current++] = new Double(train.numInstances());
result[current++] = new Double(eval.numInstances());
result[current++] = new Double(eval.meanAbsoluteError());
result[current++] = new Double(eval.rootMeanSquaredError());
result[current++] = new Double(eval.relativeAbsoluteError());
result[current++] = new Double(eval.rootRelativeSquaredError());
result[current++] = new Double(eval.correlationCoefficient());
result[current++] = new Double(eval.SFPriorEntropy());
result[current++] = new Double(eval.SFSchemeEntropy());
result[current++] = new Double(eval.SFEntropyGain());
result[current++] = new Double(eval.SFMeanPriorEntropy());
result[current++] = new Double(eval.SFMeanSchemeEntropy());
result[current++] = new Double(eval.SFMeanEntropyGain());
// Timing stats
result[current++] = new Double(trainTimeElapsed / 1000.0);
result[current++] = new Double(testTimeElapsed / 1000.0);
if (canMeasureCPUTime) {
result[current++] = new Double((trainCPUTimeElapsed / 1000000.0) / 1000.0);
result[current++] = new Double((testCPUTimeElapsed / 1000000.0) / 1000.0);
} else {
result[current++] = new Double(Utils.missingValue());
result[current++] = new Double(Utils.missingValue());
}
// sizes
if (m_NoSizeDetermination) {
result[current++] = -1.0;
result[current++] = -1.0;
result[current++] = -1.0;
} else {
ByteArrayOutputStream bastream = new ByteArrayOutputStream();
ObjectOutputStream oostream = new ObjectOutputStream(bastream);
oostream.writeObject(m_Classifier);
result[current++] = new Double(bastream.size());
bastream = new ByteArrayOutputStream();
oostream = new ObjectOutputStream(bastream);
oostream.writeObject(train);
result[current++] = new Double(bastream.size());
bastream = new ByteArrayOutputStream();
oostream = new ObjectOutputStream(bastream);
oostream.writeObject(test);
result[current++] = new Double(bastream.size());
}
// Prediction interval statistics
result[current++] = new Double(eval.coverageOfTestCasesByPredictedRegions());
result[current++] = new Double(eval.sizeOfPredictedRegions());
if (m_Classifier instanceof Summarizable) {
result[current++] = ((Summarizable) m_Classifier).toSummaryString();
} else {
result[current++] = null;
}
for (int i = 0; i < addm; i++) {
if (m_doesProduce[i]) {
try {
double dv =
((AdditionalMeasureProducer) m_Classifier).getMeasure(m_AdditionalMeasures[i]);
if (!Utils.isMissingValue(dv)) {
Double value = new Double(dv);
result[current++] = value;
} else {
result[current++] = null;
}
} catch (Exception ex) {
System.err.println(ex);
}
} else {
result[current++] = null;
}
}
// get the actual metrics from the evaluation object
List<AbstractEvaluationMetric> metrics = eval.getPluginMetrics();
if (metrics != null) {
for (AbstractEvaluationMetric m : metrics) {
if (m.appliesToNumericClass()) {
List<String> statNames = m.getStatisticNames();
for (String s : statNames) {
result[current++] = new Double(m.getStatistic(s));
}
}
}
}
if (current != RESULT_SIZE + addm + m_numPluginStatistics) {
throw new Error("Results didn't fit RESULT_SIZE");
}
return result;
}