Last version

Author: boididou

src/gr/iti/mklab/verify/Bagging.java

@@ -9,349 +9,14 @@
 import java.util.Random;
 
 import weka.classifiers.Classifier;
-import weka.classifiers.functions.LinearRegression;
 import weka.classifiers.meta.FilteredClassifier;
 import weka.classifiers.trees.RandomForest;
-import weka.core.Attribute;
 import weka.core.Instance;
 import weka.core.Instances;
-import weka.filters.unsupervised.attribute.Normalize;
 import weka.filters.unsupervised.attribute.Remove;
 
 public class Bagging {
 
-	static Normalize normFilter = new Normalize();
-	static Normalize normFilterUser = new Normalize();
-
-	static FilteredClassifier model = new FilteredClassifier();
-	static FilteredClassifier model2 = new FilteredClassifier();
-	static FilteredClassifier model3 = new FilteredClassifier();
-	static FilteredClassifier model4 = new FilteredClassifier();
-	static FilteredClassifier model5 = new FilteredClassifier();
-	static FilteredClassifier model6 = new FilteredClassifier();
-	
-	/**
-	 * Applies linear regression to the missing values of the training set in order to predict their values
-	 * and normalization to make the values fall into the range [0,1].
-	 * It is the case of the Item type Instances.
-	 * @param trainingSet
-	 * @return Instances the transformed training set
-	 * @throws Exception
-	 */
-	public static Instances getTransformedTraining(Instances trainingSet) throws Exception {
-		
-		Remove rm = new Remove();
-		rm.setAttributeIndices("1");
-			
-		ArrayList<Attribute> fvAttributes = ItemClassifier.getFvAttributes();
-		
-		//REGRESSION
-		//wotTrust		
-		trainingSet.setClass(fvAttributes.get(22)); 
-		LinearRegression lr = new LinearRegression();
-		Instances training_regr = null;
-		model.setFilter(rm);
-		model.setClassifier(lr);
-		
-		try {
-			model.buildClassifier(trainingSet);
-			training_regr = DataHandler.getInstance().applyRegressionModel(trainingSet, fvAttributes, model);
-		}
-		catch(Exception e) {
-			training_regr = trainingSet;
-			System.out.println("not enought training instances. Linear Regression not performed!");
-		}
-		
-		//readability
-		training_regr.setClass(fvAttributes.get(25));
-		LinearRegression lr2 = new LinearRegression();
-		Instances training_regr2 = null;
-		
-		model2.setFilter(rm);
-		model2.setClassifier(lr2);
-		try {
-			model2.buildClassifier(training_regr);
-			training_regr2 = DataHandler.getInstance().applyRegressionModel(training_regr, fvAttributes, model2);
-		}
-		catch(Exception e){
-			training_regr2 = training_regr;
-			System.out.println("not enought training instances. Linear Regression not performed!");
-		}
-		
-		training_regr2.setClass(fvAttributes.get(19));
-		LinearRegression lr3 = new LinearRegression();
-		Instances training_regr3 = null;
-		
-		model3.setFilter(rm);
-		model3.setClassifier(lr3);
-		try {
-			model3.buildClassifier(training_regr2);
-			training_regr3 = DataHandler.getInstance().applyRegressionModel(training_regr2, fvAttributes, model3);
-		}
-		catch(Exception e) {
-			training_regr3 = training_regr2;
-			System.out.println("not enought training instances. Linear Regression not performed!");
-		}
-		
-		training_regr3.setClass(fvAttributes.get(13));
-		LinearRegression lr4 = new LinearRegression();
-		Instances training_regr4 = null;
-		model4.setFilter(rm);
-		model4.setClassifier(lr4);
-		try {
-			model4.buildClassifier(training_regr3);
-			training_regr4 = DataHandler.getInstance().applyRegressionModel(training_regr3, fvAttributes, model4);
-		}
-		catch(Exception e) {
-			training_regr4 = training_regr3;
-			System.out.println("not enought training instances. Linear Regression not performed!");
-		}
-		
-		training_regr4.setClass(fvAttributes.get(14));
-		LinearRegression lr5 = new LinearRegression();
-		Instances training_regr5 = null;
-		model5.setFilter(rm);
-		model5.setClassifier(lr5);
-		
-		try {
-			model5.buildClassifier(training_regr4);
-			training_regr5 = DataHandler.getInstance().applyRegressionModel(training_regr4, fvAttributes, model5);
-		}
-		catch(Exception e) {
-			training_regr5 = training_regr4;
-			System.out.println("not enought training instances. Linear Regression not performed!");
-		}
-		
-		training_regr5.setClass(fvAttributes.get(6));
-		LinearRegression lr6 = new LinearRegression();
-		Instances training_regr6 = null;
-		model6.setFilter(rm);
-		model6.setClassifier(lr6);
-		try {
-			model6.buildClassifier(training_regr5);
-			training_regr6 = DataHandler.getInstance().applyRegressionModel(training_regr5, fvAttributes, model6);
-		}
-		catch(Exception e) {
-			training_regr6 = training_regr5;
-		}
-		
-						
-		// normalization part
-		String[] options = { "-S", "2.0", "-T", "-1.0" };
-		normFilter.setOptions(options);
-		normFilter.setInputFormat(training_regr5);
-		
-		Instances trainingSet_normed = DataHandler.getInstance().normalizeData(training_regr6, fvAttributes.size() - 1, normFilter);
-		
-				
-		return trainingSet_normed;
-		
-	}
-	
-	
-	static FilteredClassifier usermodel = new FilteredClassifier();
-	static FilteredClassifier usermodel2 = new FilteredClassifier();
-	static FilteredClassifier usermodel3 = new FilteredClassifier();
-	
-	/**
-	 * Applies linear regression to the missing values of the training set in order to predict their values
-	 * and normalization to make the values fall into the range [0,1].
-	 * It is the case of the User type Instances.
-	 * @param trainingSet
-	 * @return Instances the transformed training set
-	 * @throws Exception
-	 */
-	public static Instances getTransformedTrainingUser(Instances trainingSet) {
-		
-		ArrayList<Attribute> fvAttributes = UserClassifier.getFvAttributes();
-		
-		//remove filter in order to remove the id attribute
-		Remove rm = new Remove();
-		rm.setAttributeIndices("1");
-		
-		//regression
-		trainingSet.setClass(fvAttributes.get(11));
-		LinearRegression lr = new LinearRegression();
-		Instances training_regr = null;
-		usermodel.setFilter(rm);
-		usermodel.setClassifier(lr);
-
-		
-		try {
-			usermodel.buildClassifier(trainingSet);
-			training_regr  = DataHandler.getInstance().applyRegressionModel(trainingSet, fvAttributes, usermodel);
-		}
-		catch(Exception e) {
-			training_regr = trainingSet;
-			System.out.println("not enought training instances. Linear Regression not performed!");
-		}
-		
-		
-		training_regr.setClass(fvAttributes.get(13));
-		LinearRegression lr2 = new LinearRegression();
-		Instances training_regr2 = null;
-		usermodel2.setFilter(rm);
-		usermodel2.setClassifier(lr2);
-		try {
-		usermodel2.buildClassifier(training_regr); 
-			training_regr2  = DataHandler.getInstance().applyRegressionModel(training_regr, fvAttributes, usermodel2);
-		}
-		catch(Exception e) {
-			training_regr2 = training_regr;
-			System.out.println("not enought training instances. Linear Regression not performed!");
-		}
-		
-		training_regr2.setClass(fvAttributes.get(16));
-		LinearRegression lr3 = new LinearRegression();
-		Instances training_regr3 = null;
-		usermodel3.setFilter(rm);
-		usermodel3.setClassifier(lr3);
-		try {
-			usermodel3.buildClassifier(training_regr2); 
-			training_regr3  = DataHandler.getInstance().applyRegressionModel(training_regr2, fvAttributes, usermodel3);
-		}
-		catch(Exception e) {
-			training_regr3 = training_regr2;
-			System.out.println("not enought training instances. Linear Regression not performed!");
-		}
-		
-		//normalization
-		normFilterUser  =  DataHandler.getInstance().createNormalizationFilter(training_regr3);
-		Instances trainingSet_normed = DataHandler.getInstance().normalizeData(training_regr3, fvAttributes.size()-1, normFilterUser);
-				
-		return trainingSet_normed;
-	}
-	
-	/**
-	 * Applies Linear Regression and normalization to the testing set according to the models and filters
-	 * created before by the training set.
-	 * It is the case of the Item type Instances. 
-	 * @param testing the Instances to be transformed
-	 * @return the transformed testing set
-	 * @throws Exception
-	 */
-	public static Instances getTransformedTesting(Instances testing) throws Exception {
-		
-		ArrayList<Attribute> fvAttributes = ItemClassifier.getFvAttributes();
-		Instances testing_regr=null,testing_regr2=null,testing_regr3=null,testing_regr4=null,testing_regr5=null,testing_regr6=null;
-		
-		// regression
-		if (!model.toString().contains("No model built yet.")) {
-			testing.setClass(fvAttributes.get(22));
-			testing_regr = DataHandler.getInstance().applyRegressionModel(testing, fvAttributes, model);
-		}
-		else {
-			testing_regr = testing;
-		}
-		
-		if (!model2.toString().contains("No model built yet.")) {
-			testing_regr.setClass(fvAttributes.get(25));
-			testing_regr2 = DataHandler.getInstance().applyRegressionModel(testing_regr, fvAttributes, model2);
-		}
-		else {
-			testing_regr2 = testing_regr;
-		}
-		
-		if (!model3.toString().contains("No model built yet.")) {
-			testing_regr2.setClass(fvAttributes.get(19));
-			testing_regr3 = DataHandler.getInstance().applyRegressionModel(testing_regr2, fvAttributes, model3);
-		}
-		else {
-			testing_regr3 = testing_regr2;
-		}
-		
-		if (!model4.toString().contains("No model built yet.")) {
-			testing_regr3.setClass(fvAttributes.get(13));
-			testing_regr4 = DataHandler.getInstance().applyRegressionModel(testing_regr3, fvAttributes, model4);
-		}
-		else {
-			testing_regr4 = testing_regr3;
-		}
-		
-		if (!model5.toString().contains("No model built yet.")) {
-			testing_regr4.setClass(fvAttributes.get(14));
-			testing_regr5 = DataHandler.getInstance().applyRegressionModel(testing_regr4, fvAttributes, model5);
-		}
-		else {
-			testing_regr5 = testing_regr4;
-		}
-		
-		if (!model6.toString().contains("No model built yet.")) {
-			testing_regr5.setClass(fvAttributes.get(6));
-			testing_regr6 = DataHandler.getInstance().applyRegressionModel(testing_regr5, fvAttributes, model6);
-		}
-		else {
-			testing_regr6 = testing_regr5;
-		}
-			
-
-		// normalization
-		Instances testingSet_normed = DataHandler.getInstance().normalizeData(testing_regr6, fvAttributes.size() - 1, normFilter);
-		//testingSet_normed = getTrimmedInstances(testingSet_normed);
-
-		return testingSet_normed;
-	}
-	
-	
-	/**
-	 * Applies Linear Regression and normalization to the testing set according to the models and filters
-	 * created before by the training set.
-	 * It is the case of the User type Instances. 
-	 * @param testing the Instances to be transformed
-	 * @return the transformed testing set
-	 * @throws Exception
-	 */
-	public static Instances getTransformedTestingUser(Instances testing) throws Exception {
-		
-		ArrayList<Attribute> fvAttributes = UserClassifier.getFvAttributes();
-		Instances testing_regr=null,testing_regr2=null,testing_regr3=null;
-		
-		testing_regr3 = testing;
-		
-		if (!usermodel.toString().contains("No model built yet.")) {
-			testing.setClass(fvAttributes.get(11));
-			testing_regr  = DataHandler.getInstance().applyRegressionModel(testing, fvAttributes, usermodel);
-		}
-		else {
-			testing_regr = testing;
-		}
-		if (!usermodel2.toString().contains("No model built yet.")) {
-			testing_regr.setClass(fvAttributes.get(13));
-			testing_regr2  = DataHandler.getInstance().applyRegressionModel(testing_regr, fvAttributes, usermodel2);
-		}
-		else {
-			testing_regr2= testing_regr;
-		}
-		if (!usermodel3.toString().contains("No model built yet.")) {
-			testing_regr2.setClass(fvAttributes.get(16));
-			testing_regr3  = DataHandler.getInstance().applyRegressionModel(testing_regr2, fvAttributes, usermodel3);
-		}
-		else {
-			testing_regr3 = testing_regr2;
-		}
-		
-		if (!usermodel.toString().contains("No model built yet.")) {
-			testing.setClass(fvAttributes.get(1));
-			testing_regr  = DataHandler.getInstance().applyRegressionModel(testing, fvAttributes, usermodel);
-		}
-		else {
-			testing_regr = testing;
-		}
-		if (!usermodel2.toString().contains("No model built yet.")) {
-			testing_regr.setClass(fvAttributes.get(4));
-			testing_regr2  = DataHandler.getInstance().applyRegressionModel(testing_regr, fvAttributes, usermodel2);
-		}
-		else {
-			testing_regr2= testing_regr;
-		}
-		
-		
-		//normalization
-		Instances testingSet_normed = DataHandler.getInstance().normalizeData(testing_regr3, fvAttributes.size()-1, normFilterUser);
-		//testingSet_normed = getTrimmedInstances(testingSet_normed);
-		
-		return testingSet_normed;
-	}
 
 	static Instances[] testingSets = new Instances[9];