// 构造一个tri-trainer分类器。
public Tritrainer(
String classifier, String trainingIns_File, String testIns_File, double precentage) {
try {
this.classifier1 = (Classifier) Class.forName(classifier).newInstance();
this.classifier2 = (Classifier) Class.forName(classifier).newInstance();
this.classifier3 = (Classifier) Class.forName(classifier).newInstance();
Instances trainingInstances = Util.getInstances(trainingIns_File);
// 将trainIns_File按照precentage和(1-precentage)的比例切割成labeledIns和unlabeledIns;
int length = trainingInstances.numInstances();
int i = new Double(length * precentage).intValue();
labeledIns = new Instances(trainingInstances, 0);
for (int j = 0; j < i; j++) {
labeledIns.add(trainingInstances.firstInstance());
trainingInstances.delete(0);
}
unlabeledIns = trainingInstances;
testIns = Util.getInstances(testIns_File);
Init();
} catch (Exception e) {
}
}
// 将样本集中裁剪提取成m个样本组成的集合；
public void SubSample(Instances inst, int m) {
inst.randomize(new Random());
while (inst.numInstances() != m) {
inst.delete(0);
}
// System.out.println("subsample:=" + inst.numInstances() + " m:=" + m );
}
/**
* This will remove all buffered instances from the inputformat dataset. Use this method rather
* than getInputFormat().delete();
*/
protected void flushInput() {
if ((m_InputStringAtts.getAttributeIndices().length > 0)
|| (m_InputRelAtts.getAttributeIndices().length > 0)) {
m_InputFormat = m_InputFormat.stringFreeStructure();
m_InputStringAtts = new StringLocator(m_InputFormat, m_InputStringAtts.getAllowedIndices());
m_InputRelAtts = new RelationalLocator(m_InputFormat, m_InputRelAtts.getAllowedIndices());
} else {
// This more efficient than new Instances(m_InputFormat, 0);
m_InputFormat.delete();
}
}
/**
* Process the input and compute the statistics of the training data
*
* @return an Input object which holds the statistics about the training data
*/
private Input sIB_ProcessInput() {
double valSum = 0.0;
for (int i = 0; i < m_numInstances; i++) {
valSum = 0.0;
for (int v = 0; v < m_data.instance(i).numValues(); v++) {
valSum += m_data.instance(i).valueSparse(v);
}
if (valSum <= 0) {
if (m_verbose) {
System.out.format("Instance %s sum of value = %s <= 0, removed.\n", i, valSum);
}
m_data.delete(i);
m_numInstances--;
}
}
// get the term-document matrix
Input input = new Input();
input.Py_x = getTransposedNormedMatrix(m_data);
if (m_uniformPrior) {
input.Pyx = input.Py_x.copy();
normalizePrior(m_data);
} else {
input.Pyx = getTransposedMatrix(m_data);
}
input.sumVals = getTotalSum(m_data);
input.Pyx.timesEquals(1 / input.sumVals);
// prior probability of documents, ie. sum the columns from the Pyx matrix
input.Px = new double[m_numInstances];
for (int i = 0; i < m_numInstances; i++) {
for (int j = 0; j < m_numAttributes; j++) {
input.Px[i] += input.Pyx.get(j, i);
}
}
// prior probability of terms, ie. sum the rows from the Pyx matrix
input.Py = new double[m_numAttributes];
for (int i = 0; i < input.Pyx.getRowDimension(); i++) {
for (int j = 0; j < input.Pyx.getColumnDimension(); j++) {
input.Py[i] += input.Pyx.get(i, j);
}
}
MI(input.Pyx, input);
return input;
}
/**
* Read the data set incrementally---get the next instance in the data set or returns null if
* there are no more instances to get. If the structure hasn't yet been determined by a call to
* getStructure then method should do so before returning the next instance in the data set.
*
* @return the next instance in the data set as an Instance object or null if there are no more
* instances to be read
* @exception IOException if there is an error during parsing
*/
public Instance getNextInstance() throws IOException {
if (m_structure == null) {
getStructure();
}
if (getRetrieval() == BATCH) {
throw new IOException("Cannot mix getting Instances in both incremental and batch modes");
}
setRetrieval(INCREMENTAL);
if (!m_structure.readInstance(m_sourceReader)) {
return null;
}
Instance current = m_structure.instance(0);
m_structure.delete(0);
return current;
}
/**
* Adds the supplied instance to the training set.
*
* @param instance the instance to add
* @throws Exception if instance could not be incorporated successfully
*/
public void updateClassifier(Instance instance) throws Exception {
if (m_Train.equalHeaders(instance.dataset()) == false) {
throw new Exception("Incompatible instance types");
}
if (instance.classIsMissing()) {
return;
}
m_Train.add(instance);
m_NNSearch.update(instance);
m_kNNValid = false;
if ((m_WindowSize > 0) && (m_Train.numInstances() > m_WindowSize)) {
boolean deletedInstance = false;
while (m_Train.numInstances() > m_WindowSize) {
m_Train.delete(0);
deletedInstance = true;
}
// rebuild datastructure KDTree currently can't delete
if (deletedInstance == true) m_NNSearch.setInstances(m_Train);
}
}
/**
* Process an instance. When the batch size is met then a batch of predictions is returned;
* returns null if the batch size has not been matched yet
*
* @param inst the instance to process
* @return a batch of predictions or null if we have not seen enough input instances to meet the
* batch size yet
* @throws DistributedWekaException if a problem occurs
*/
public double[][] processInstanceBatchPredictor(Instance inst) throws DistributedWekaException {
inst = mapIncomingFieldsToModelFields(inst);
m_batchScoringData.add(inst);
if (m_batchScoringData.numInstances() == m_batchSize) {
try {
double[][] predictions = m_model.distributionsForInstances(m_batchScoringData);
if (predictions.length != m_batchScoringData.numInstances()) {
throw new Exception(
"Number of predictions did not match the number " + "of instances in the batch");
}
m_batchScoringData.delete();
return predictions;
} catch (Exception ex) {
throw new DistributedWekaException(ex);
}
}
return null;
}
public static void execSVM(String expName) {
try {
FileWriter outFile = null;
PrintWriter out = null;
outFile = new FileWriter(expName + "-SVM.results");
out = new PrintWriter(outFile);
DateFormat dateFormat = new SimpleDateFormat("yyyy/MM/dd HH:mm:ss");
ProcessTweets tweetsProcessor = null;
System.out.println("***************************************");
System.out.println("***\tEXECUTING TEST\t" + expName + "***");
System.out.println("+++++++++++++++++++++++++++++++++++++++");
out.println("***************************************");
out.println("***\tEXECUTING TEST\t" + expName + "***");
out.println("+++++++++++++++++++++++++++++++++++++++");
out.println("4-Generate classifier " + dateFormat.format(new Date()));
Classifier cls = null;
DataSource sourceTrain = new DataSource(expName + "-train.arff");
Instances dataTrain = sourceTrain.getDataSet();
if (dataTrain.classIndex() == -1) dataTrain.setClassIndex(dataTrain.numAttributes() - 1);
// Entreno el clasificador
// cls = new weka.classifiers.functions.LibSVM();
int clase = dataTrain.numAttributes() - 1;
cls = new weka.classifiers.bayes.ComplementNaiveBayes();
dataTrain.setClassIndex(clase);
cls.buildClassifier(dataTrain);
ObjectOutputStream oos =
new ObjectOutputStream(new FileOutputStream(expName + "-SVM.classifier"));
oos.writeObject(cls);
oos.flush();
oos.close();
DataSource sourceTest = new DataSource(expName + "-test.arff");
Instances dataTest = sourceTest.getDataSet();
dataTest.setClassIndex(clase);
Evaluation eval = new Evaluation(dataTest);
eval.evaluateModel(cls, dataTest);
// Ahora calculo los valores precision, recall y fmeasure. Además saco las matrices de
// confusion
float precision = 0;
float recall = 0;
float fmeasure = 0;
int topeTopics = 8;
for (int ind = 0; ind < topeTopics; ind++) {
precision += eval.precision(ind);
recall += eval.recall(ind);
fmeasure += eval.fMeasure(ind);
}
precision = precision / topeTopics;
recall = recall / topeTopics;
fmeasure = fmeasure / topeTopics;
System.out.println("++++++++++++++ CNB ++++++++++++++++++++");
System.out.println(eval.toMatrixString());
System.out.println("+++++++++++++++++++++++++++++++++++++++");
System.out.printf("Precision: %.3f\n", precision);
System.out.printf("Recall: %.3f\n", recall);
System.out.printf("F-measure: %.3f\n", fmeasure);
System.out.println("***************************************");
out.println("++++++++++++++ CNB ++++++++++++++++++++");
out.println(eval.toMatrixString());
out.println("+++++++++++++++++++++++++++++++++++++++");
out.printf("Precision: %.3f\n", precision);
out.printf("Recall: %.3f\n", recall);
out.printf("F-measure: %.3f\n", fmeasure);
out.println("***************************************");
// OTRO CLASIFICADOR ZeroR
cls = new weka.classifiers.rules.ZeroR();
dataTrain.setClassIndex(clase);
cls.buildClassifier(dataTrain);
eval = new Evaluation(dataTest);
eval.evaluateModel(cls, dataTest);
precision = 0;
recall = 0;
fmeasure = 0;
for (int ind = 0; ind < topeTopics; ind++) {
precision += eval.precision(ind);
recall += eval.recall(ind);
fmeasure += eval.fMeasure(ind);
}
precision = precision / topeTopics;
recall = recall / topeTopics;
fmeasure = fmeasure / topeTopics;
System.out.println("++++++++++++++ ZEROR ++++++++++++++++++++");
System.out.println(eval.toMatrixString());
System.out.println("+++++++++++++++++++++++++++++++++++++++");
System.out.printf("Precision: %.3f\n", precision);
System.out.printf("Recall: %.3f\n", recall);
System.out.printf("F-measure: %.3f\n", fmeasure);
System.out.println("***************************************");
out.println("++++++++++++++ ZEROR ++++++++++++++++++++");
out.println(eval.toMatrixString());
out.println("+++++++++++++++++++++++++++++++++++++++");
out.printf("Precision: %.3f\n", precision);
out.printf("Recall: %.3f\n", recall);
out.printf("F-measure: %.3f\n", fmeasure);
out.println("***************************************");
// OTRO CLASIFICADOR J48
/*
cls = new weka.classifiers.trees.J48();
dataTrain.setClassIndex(clase);
cls.buildClassifier(dataTrain);
eval = new Evaluation(dataTest);
eval.evaluateModel(cls, dataTest);
precision=0;
recall=0;
fmeasure=0;
for(int ind=0; ind<topeTopics; ind++)
{
precision += eval.precision(ind);
recall += eval.recall(ind);
fmeasure += eval.fMeasure(ind);
}
precision = precision / topeTopics;
recall = recall / topeTopics;
fmeasure = fmeasure / topeTopics;
System.out.println("++++++++++++++ J48 ++++++++++++++++++++");
System.out.println(eval.toMatrixString());
System.out.println("+++++++++++++++++++++++++++++++++++++++");
System.out.printf("Precision: %.3f\n", precision);
System.out.printf("Recall: %.3f\n", recall);
System.out.printf("F-measure: %.3f\n", fmeasure);
System.out.println("***************************************");
out.println("++++++++++++++ J48 ++++++++++++++++++++");
out.println(eval.toMatrixString());
out.println("+++++++++++++++++++++++++++++++++++++++");
out.printf("Precision: %.3f\n", precision);
out.printf("Recall: %.3f\n", recall);
out.printf("F-measure: %.3f\n", fmeasure);
out.println("***************************************");
//OTRO SMO
cls = new weka.classifiers.functions.SMO();
dataTrain.setClassIndex(clase);
cls.buildClassifier(dataTrain);
eval = new Evaluation(dataTest);
eval.evaluateModel(cls, dataTest);
precision=0;
recall=0;
fmeasure=0;
for(int ind=0; ind<topeTopics; ind++)
{
precision += eval.precision(ind);
recall += eval.recall(ind);
fmeasure += eval.fMeasure(ind);
}
precision = precision / topeTopics;
recall = recall / topeTopics;
fmeasure = fmeasure / topeTopics;
System.out.println("++++++++++++++ SMO ++++++++++++++++++++");
System.out.println(eval.toMatrixString());
System.out.println("+++++++++++++++++++++++++++++++++++++++");
System.out.printf("Precision: %.3f\n", precision);
System.out.printf("Recall: %.3f\n", recall);
System.out.printf("F-measure: %.3f\n", fmeasure);
System.out.println("***************************************");
out.println("++++++++++++++ SMO ++++++++++++++++++++");
out.println(eval.toMatrixString());
out.println("+++++++++++++++++++++++++++++++++++++++");
out.printf("Precision: %.3f\n", precision);
out.printf("Recall: %.3f\n", recall);
out.printf("F-measure: %.3f\n", fmeasure);
out.println("***************************************");
*/
out.flush();
out.close();
dataTest.delete();
dataTrain.delete();
} catch (FileNotFoundException e) {
// TODO Auto-generated catch block
e.printStackTrace();
} catch (IOException e) {
// TODO Auto-generated catch block
e.printStackTrace();
} catch (Exception e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
}
public void exec(PrintWriter printer) {
try {
FileWriter outFile = null;
PrintWriter out = null;
if (printer == null) {
outFile = new FileWriter(id + ".results");
out = new PrintWriter(outFile);
} else out = printer;
DateFormat dateFormat = new SimpleDateFormat("yyyy/MM/dd HH:mm:ss");
ProcessTweets tweetsProcessor = null;
System.out.println("***************************************");
System.out.println("***\tEXECUTING TEST\t" + id + "***");
System.out.println("+++++++++++++++++++++++++++++++++++++++");
System.out.println("Train size:" + traincorpus.size());
System.out.println("Test size:" + testcorpus.size());
out.println("***************************************");
out.println("***\tEXECUTING TEST\t***");
out.println("+++++++++++++++++++++++++++++++++++++++");
out.println("Train size:" + traincorpus.size());
out.println("Test size:" + testcorpus.size());
String cloneID = "";
boolean clonar = false;
if (baseline) {
System.out.println("***************************************");
System.out.println("***\tEXECUTING TEST BASELINE\t***");
System.out.println("+++++++++++++++++++++++++++++++++++++++");
System.out.println("Train size:" + traincorpus.size());
System.out.println("Test size:" + testcorpus.size());
out.println("***************************************");
out.println("***\tEXECUTING TEST\t***");
out.println("+++++++++++++++++++++++++++++++++++++++");
out.println("Train size:" + traincorpus.size());
out.println("Test size:" + testcorpus.size());
BaselineClassifier base = new BaselineClassifier(testcorpus, 8);
precision = base.getPrecision();
recall = base.getRecall();
fmeasure = base.getFmeasure();
System.out.println("+++++++++++++++++++++++++++++++++++++++");
System.out.printf("Precision: %.3f\n", precision);
System.out.printf("Recall: %.3f\n", recall);
System.out.printf("F-measure: %.3f\n", fmeasure);
System.out.println("***************************************");
out.println("+++++++++++++++++++++++++++++++++++++++");
out.printf("Precision: %.3f\n", precision);
out.printf("Recall: %.3f\n", recall);
out.printf("F-measure: %.3f\n", fmeasure);
out.println("***************************************");
out.flush();
out.close();
return;
} else {
System.out.println("Stemming: " + stemming);
System.out.println("Lematization:" + lematization);
System.out.println("URLs:" + urls);
System.out.println("Hashtags:" + hashtags);
System.out.println("Mentions:" + mentions);
System.out.println("Unigrams:" + unigrams);
System.out.println("Bigrams:" + bigrams);
System.out.println("TF:" + tf);
System.out.println("TF-IDF:" + tfidf);
out.println("Stemming: " + stemming);
out.println("Lematization:" + lematization);
out.println("URLs:" + urls);
out.println("Hashtags:" + hashtags);
out.println("Mentions:" + mentions);
out.println("Unigrams:" + unigrams);
out.println("Bigrams:" + bigrams);
out.println("TF:" + tf);
out.println("TF-IDF:" + tfidf);
}
// Si tengo los tweets procesados, me evito un nuevo proceso
System.out.println("1-Process tweets " + dateFormat.format(new Date()));
out.println("1-Process tweets " + dateFormat.format(new Date()));
List<ProcessedTweet> train = null;
String[] ids = id.split("-");
cloneID = ids[0] + "-" + (Integer.valueOf(ids[1]) + 6);
if (((Integer.valueOf(ids[1]) / 6) % 2) == 0) clonar = true;
if (new File(id + "-train.ptweets").exists()) {
train = ProcessedTweetSerialization.fromFile(id + "-train.ptweets");
tweetsProcessor =
new ProcessTweets(stemming, lematization, urls, hashtags, mentions, unigrams, bigrams);
if (lematization) {
tweetsProcessor.doLematization(train);
}
if (stemming) {
tweetsProcessor.doStemming(train);
}
} else {
tweetsProcessor =
new ProcessTweets(stemming, lematization, urls, hashtags, mentions, unigrams, bigrams);
// Esto del set training es un añadido para poder diferenciar los idiomas de las url en el
// corpus paralelo
// tweetsProcessor.setTraining(true);
train = tweetsProcessor.processTweets(traincorpus);
// tweetsProcessor.setTraining(false);
ProcessedTweetSerialization.toFile(id + "-train.ptweets", train);
/*
if (clonar)
{
File f = new File (id+"-train.ptweets");
Path p = f.toPath();
CopyOption[] options = new CopyOption[]{
StandardCopyOption.REPLACE_EXISTING,
StandardCopyOption.COPY_ATTRIBUTES
};
Files.copy(p, new File (cloneID+"-train.ptweets").toPath(), options);
Files.copy(p, new File (ids[0]+"-"+(Integer.valueOf(ids[1])+12)+"-train.ptweets").toPath(), options);
Files.copy(p, new File (ids[0]+"-"+(Integer.valueOf(ids[1])+18)+"-train.ptweets").toPath(), options);
Files.copy(p, new File (ids[0]+"-"+(Integer.valueOf(ids[1])+24)+"-train.ptweets").toPath(), options);
Files.copy(p, new File (ids[0]+"-"+(Integer.valueOf(ids[1])+30)+"-train.ptweets").toPath(), options);
}
*/
}
// Generamos las BOW. Igual que antes, si existen no las creo.
System.out.println("2-Fill topics " + dateFormat.format(new Date()));
out.println("2-Fill topics " + dateFormat.format(new Date()));
TopicsList topics = null;
if (new File(id + ".topics").exists()) {
topics = TopicsSerialization.fromFile(id + ".topics");
if (tf) topics.setSelectionFeature(TopicDesc.TERM_TF);
else topics.setSelectionFeature(TopicDesc.TERM_TF_IDF);
topics.prepareTopics();
} else {
topics = new TopicsList();
if (tf) topics.setSelectionFeature(TopicDesc.TERM_TF);
else topics.setSelectionFeature(TopicDesc.TERM_TF_IDF);
System.out.println("Filling topics " + dateFormat.format(new Date()));
topics.fillTopics(train);
System.out.println("Preparing topics topics " + dateFormat.format(new Date()));
// Aquí tengo que serializar antes de preparar, porque si no no puedo calcular los tf y
// tfidf
System.out.println("Serializing topics topics " + dateFormat.format(new Date()));
/*
if (clonar)
{
TopicsSerialization.toFile(cloneID+".topics", topics);
}
*/
topics.prepareTopics();
TopicsSerialization.toFile(id + ".topics", topics);
}
System.out.println("3-Generate arff train file " + dateFormat.format(new Date()));
out.println("3-Generate arff train file " + dateFormat.format(new Date()));
// Si el fichero arff no existe, lo creo. en caso contrario vengo haciendo lo que hasta ahora,
// aprovechar trabajo previo
if (!new File(id + "-train.arff").exists()) {
BufferedWriter bw = topics.generateArffHeader(id + "-train.arff");
int tope = traincorpus.size();
if (tweetsProcessor == null)
tweetsProcessor =
new ProcessTweets(
stemming, lematization, urls, hashtags, mentions, unigrams, bigrams);
for (int indTweet = 0; indTweet < tope; indTweet++) {
topics.generateArffVector(bw, train.get(indTweet));
}
bw.flush();
bw.close();
}
// Ahora proceso los datos de test
System.out.println("5-build test dataset " + dateFormat.format(new Date()));
out.println("5-build test dataset " + dateFormat.format(new Date()));
List<ProcessedTweet> test = null;
if (new File(id + "-test.ptweets").exists())
test = ProcessedTweetSerialization.fromFile(id + "-test.ptweets");
else {
if (tweetsProcessor == null)
tweetsProcessor =
new ProcessTweets(
stemming, lematization, urls, hashtags, mentions, unigrams, bigrams);
test = tweetsProcessor.processTweets(testcorpus);
ProcessedTweetSerialization.toFile(id + "-test.ptweets", test);
/*
if (clonar)
{
File f = new File (id+"-test.ptweets");
Path p = f.toPath();
CopyOption[] options = new CopyOption[]{
StandardCopyOption.REPLACE_EXISTING,
StandardCopyOption.COPY_ATTRIBUTES
};
Files.copy(p, new File (cloneID+"-test.ptweets").toPath(), options);
}
*/
}
// Si el fichero arff no existe, lo creo. en caso contrario vengo haciendo lo que hasta ahora,
// aprovechar trabajo previo
if (!new File(id + "-test.arff").exists()) {
BufferedWriter bw = topics.generateArffHeader(id + "-test.arff");
int tope = testcorpus.size();
if (tweetsProcessor == null)
tweetsProcessor =
new ProcessTweets(
stemming, lematization, urls, hashtags, mentions, unigrams, bigrams);
for (int indTweet = 0; indTweet < tope; indTweet++) {
topics.generateArffVector(bw, test.get(indTweet));
}
bw.flush();
bw.close();
}
int topeTopics = topics.getTopicsList().size();
topics.getTopicsList().clear();
// Genero el clasificador
// FJRM 25-08-2013 Lo cambio de orden para intentar liberar la memoria de los topics y tener
// más libre
System.out.println("4-Generate classifier " + dateFormat.format(new Date()));
out.println("4-Generate classifier " + dateFormat.format(new Date()));
Classifier cls = null;
DataSource sourceTrain = null;
Instances dataTrain = null;
if (new File(id + "-MNB.classifier").exists()) {
ObjectInputStream ois = new ObjectInputStream(new FileInputStream(id + "-MNB.classifier"));
cls = (Classifier) ois.readObject();
ois.close();
} else {
sourceTrain = new DataSource(id + "-train.arff");
dataTrain = sourceTrain.getDataSet();
if (dataTrain.classIndex() == -1) dataTrain.setClassIndex(dataTrain.numAttributes() - 1);
// Entreno el clasificador
cls = new weka.classifiers.bayes.NaiveBayesMultinomial();
int clase = dataTrain.numAttributes() - 1;
dataTrain.setClassIndex(clase);
cls.buildClassifier(dataTrain);
ObjectOutputStream oos =
new ObjectOutputStream(new FileOutputStream(id + "-MNB.classifier"));
oos.writeObject(cls);
oos.flush();
oos.close();
// data.delete();//no borro para el svm
}
// Ahora evaluo el clasificador con los datos de test
System.out.println("6-Evaluate classifier MNB " + dateFormat.format(new Date()));
out.println("6-Evaluate classifier MNB" + dateFormat.format(new Date()));
DataSource sourceTest = new DataSource(id + "-test.arff");
Instances dataTest = sourceTest.getDataSet();
int clase = dataTest.numAttributes() - 1;
dataTest.setClassIndex(clase);
Evaluation eval = new Evaluation(dataTest);
eval.evaluateModel(cls, dataTest);
// Ahora calculo los valores precision, recall y fmeasure. Además saco las matrices de
// confusion
precision = 0;
recall = 0;
fmeasure = 0;
for (int ind = 0; ind < topeTopics; ind++) {
precision += eval.precision(ind);
recall += eval.recall(ind);
fmeasure += eval.fMeasure(ind);
}
precision = precision / topeTopics;
recall = recall / topeTopics;
fmeasure = fmeasure / topeTopics;
System.out.println("+++++++++++++++++++++++++++++++++++++++");
System.out.println(eval.toMatrixString());
System.out.println("+++++++++++++++++++++++++++++++++++++++");
System.out.printf("Precision: %.3f\n", precision);
System.out.printf("Recall: %.3f\n", recall);
System.out.printf("F-measure: %.3f\n", fmeasure);
System.out.println("***************************************");
out.println("+++++++++++++++++++++++++++++++++++++++");
out.println(eval.toMatrixString());
out.println("+++++++++++++++++++++++++++++++++++++++");
out.printf("Precision: %.3f\n", precision);
out.printf("Recall: %.3f\n", recall);
out.printf("F-measure: %.3f\n", fmeasure);
out.println("***************************************");
/* NO BORRAR
System.out.println("7-Evaluate classifier SVM"+dateFormat.format(new Date()));
out.println("7-Evaluate classifier SVM"+dateFormat.format(new Date()));
if (new File(id+"-SVM.classifier").exists())
{
ObjectInputStream ois = new ObjectInputStream(new FileInputStream(id+"-SVM.classifier"));
cls = (Classifier) ois.readObject();
ois.close();
}
else
{
if (dataTrain==null)
{
sourceTrain = new DataSource(id+"-train.arff");
dataTrain = sourceTrain.getDataSet();
if (dataTrain.classIndex() == -1)
dataTrain.setClassIndex(dataTrain.numAttributes() - 1);
}
//Entreno el clasificador
cls = new weka.classifiers.functions.LibSVM();
clase = dataTrain.numAttributes()-1;
dataTrain.setClassIndex(clase);
cls.buildClassifier(dataTrain);
ObjectOutputStream oos = new ObjectOutputStream(new FileOutputStream(id+"-SVM.classifier"));
oos.writeObject(cls);
oos.flush();
oos.close();
dataTrain.delete();
}
eval.evaluateModel(cls, dataTest);
precision=0;
recall=0;
fmeasure=0;
for(int ind=0; ind<topeTopics; ind++)
{
precision += eval.precision(ind);
recall += eval.recall(ind);
fmeasure += eval.fMeasure(ind);
}
precision = precision / topeTopics;
recall = recall / topeTopics;
fmeasure = fmeasure / topeTopics;
System.out.println("+++++++++++++++++++++++++++++++++++++++");
System.out.println(eval.toMatrixString());
System.out.println("+++++++++++++++++++++++++++++++++++++++");
System.out.printf("Precision: %.3f\n", precision);
System.out.printf("Recall: %.3f\n", recall);
System.out.printf("F-measure: %.3f\n", fmeasure);
System.out.println("***************************************");
out.println("+++++++++++++++++++++++++++++++++++++++");
out.println(eval.toMatrixString());
out.println("+++++++++++++++++++++++++++++++++++++++");
out.printf("Precision: %.3f\n", precision);
out.printf("Recall: %.3f\n", recall);
out.printf("F-measure: %.3f\n", fmeasure);
out.println("***************************************");
*/
System.out.println("Done " + dateFormat.format(new Date()));
out.println("Done " + dateFormat.format(new Date()));
if (printer == null) {
out.flush();
out.close();
}
// Intento de liberar memoria
if (dataTrain != null) dataTrain.delete();
if (dataTest != null) dataTest.delete();
if (train != null) train.clear();
if (test != null) test.clear();
if (topics != null) {
topics.getTopicsList().clear();
topics = null;
}
if (dataTest != null) dataTest.delete();
if (cls != null) cls = null;
if (tweetsProcessor != null) tweetsProcessor = null;
System.gc();
} catch (Exception e) {
e.printStackTrace();
}
}
/**
* Removes a specified number of instances from the given set of instances.
*
* @param data given instances
* @param numRemove number of instances to delete from the given instances
*/
protected void removeInstances(Instances data, int numRemove) {
int num = data.numInstances();
for (int i = num - 1; i > num - 1 - numRemove; i--) {
data.delete(i);
}
}
/**
* Calculates the class membership probabilities for the given test instance.
*
* @param instance the instance to be classified
* @return predicted class probability distribution
* @throws Exception if an error occurred during the prediction
*/
public double[] distributionForInstance(Instance instance) throws Exception {
NaiveBayes nb = new NaiveBayes();
// System.out.println("number of instances "+m_Train.numInstances());
if (m_Train.numInstances() == 0) {
// throw new Exception("No training instances!");
return m_defaultModel.distributionForInstance(instance);
}
if ((m_WindowSize > 0) && (m_Train.numInstances() > m_WindowSize)) {
m_kNNValid = false;
boolean deletedInstance = false;
while (m_Train.numInstances() > m_WindowSize) {
m_Train.delete(0);
}
// rebuild datastructure KDTree currently can't delete
if (deletedInstance == true) m_NNSearch.setInstances(m_Train);
}
// Select k by cross validation
if (!m_kNNValid && (m_CrossValidate) && (m_kNNUpper >= 1)) {
crossValidate();
}
m_NNSearch.addInstanceInfo(instance);
m_kNN = 1000;
Instances neighbours = m_NNSearch.kNearestNeighbours(instance, m_kNN);
double[] distances = m_NNSearch.getDistances();
// System.out.println("--------------classify instance--------- ");
// System.out.println("neighbours.numInstances"+neighbours.numInstances());
// System.out.println("distances.length"+distances.length);
// System.out.println("--------------classify instance--------- ");
/* for (int k = 0; k < distances.length; k++) {
//System.out.println("-------");
//System.out.println("distance of "+k+" "+distances[k]);
//System.out.println("instance of "+k+" "+neighbours.instance(k));
//distances[k] = distances[k]+0.1;
//System.out.println("------- after add 0.1");
//System.out.println("distance of "+k+" "+distances[k]);
}
*/
Instances instances = new Instances(m_Train);
// int attrnum = instances.numAttributes();
instances.deleteWithMissingClass();
Instances newm_Train = new Instances(instances, 0, instances.numInstances());
for (int k = 0; k < neighbours.numInstances(); k++) {
// System.out.println("-------");
// Instance in = new Instance();
Instance insk = neighbours.instance(k);
// System.out.println("instance "+k+" "+neighbours.instance(k));
// System.out.println("-------");
double dis = distances[k] + 0.1;
// System.out.println("dis "+dis);
dis = (1 / dis) * 10;
// System.out.println("1/dis "+dis);
int weightnum = (int) dis;
// System.out.println("weightnum "+weightnum);
for (int s = 0; s < weightnum; s++) {
newm_Train.add(insk);
}
}
// System.out.println("number of instances "+newm_Train.numInstances());
/* for (int k = 0; k < newm_Train.numInstances(); k++) {
System.out.println("-------");
System.out.println("instance "+k+" "+newm_Train.instance(k));
System.out.println("-------");
}
/*
for (int k = 0; k < distances.length; k++) {
System.out.println("-------");
System.out.println("distance of "+k+" "+distances[k]);
System.out.println("-------");
}*/
nb.buildClassifier(newm_Train);
double[] dis = nb.distributionForInstance(instance);
// double[] distribution = makeDistribution(neighbours, distances);
return dis;
// return distribution;
}
