AuthorClassificationRelaxed.java

import weka.attributeSelection.InfoGainAttributeEval;
import weka.attributeSelection.Ranker;
import weka.classifiers.*;
import weka.classifiers.evaluation.ThresholdCurve;
import weka.classifiers.meta.FilteredClassifier;
import weka.classifiers.trees.RandomForest;
import weka.core.Attribute;
import weka.core.AttributeStats;
import weka.core.Instances;
import weka.core.Range;
import weka.core.Utils;
import weka.filters.Filter;
import weka.filters.supervised.attribute.AttributeSelection;
import weka.filters.unsupervised.attribute.Remove;
import weka.filters.unsupervised.instance.RemoveRange;
import weka.filters.unsupervised.instance.RemoveWithValues;

import java.io.BufferedWriter;
import java.io.FileReader;
import java.io.FileWriter;
import java.util.*;

public class AuthorClassificationRelaxed {


	public static void main(String[] args) throws Exception 
	{
		double accuracy=0;
		int endRelax = 2;
		int numberFiles;
		int numFeatures=0; //0 is the default logM+1
		int seedNumber;
		double total =0;
		double average =0;

		String fileName  ="/Users/Aylin/Desktop/resu.txt";
		

		
		for(int authorNo=9; authorNo<=10; authorNo++){
			for(numberFiles=9; numberFiles<=10; numberFiles++){
				for (int x=28; x<=(18*31); x=x+9){
				String arffFile = "/Users/Aylin/Desktop/Princeton/BAA/arffs/"
		    			+ "C_62Authors14files_decompiledNEW2.arff";
			
			  Util.writeFile(numberFiles+"FilesPerAuthor: \n",fileName, true);	
			  for(int relaxPar = 0; relaxPar<=endRelax; relaxPar++){
				  total=0;				  
				  average=0;

				  for(seedNumber=1; seedNumber<2; seedNumber++){
					  int foldNumber=numberFiles;


 
		RandomForest cls = new RandomForest();
		Instances data = new Instances(new FileReader(arffFile));
		data.setClassIndex(data.numAttributes() - 1);
//		data.setClassIndex(0);

		//do not stratify if you are going to remove instances for training and testing
//		data.stratify(foldNumber);
		

		//write classes that have 9 samples to a new arff
		System.out.println(data.attributeStats(0));

//		System.out.println(data.instance(2).stringValue(0));
//		System.out.println(data.instance(2).value(0));

/*		for(int i=0; i<=data.numInstances();i++){
			int count = data.attributeStats(0).nominalCounts[(int) data.instance(i).value(0)];
			if(count==9){
				Util.writeFile(data.instance(i).toString() + "\n", "/Users/Aylin/Desktop/"
			 		+ "python9files.arff", true);

			}
			
		}*/
		
		//Start information gain that selects up to 200 features that have nonzero infogain
		int n = 500; // number of features to select 
	    AttributeSelection attributeSelection = new  AttributeSelection(); 
	     Ranker ranker = new Ranker(); 
	     ranker.setNumToSelect(n); 
	     ranker.setThreshold(0.001);
	     InfoGainAttributeEval infoGainAttributeEval = new InfoGainAttributeEval(); 
	     attributeSelection.setEvaluator(infoGainAttributeEval); 
	     attributeSelection.setSearch(ranker); 
	     attributeSelection.setInputFormat(data); 
	     data = Filter.useFilter(data, attributeSelection); 
	     //end of infogain
	     
		
		
		RemoveRange rm = new RemoveRange();
		rm.setInputFormat(data);
	//	rm.setInstancesIndices("first-"+(x-19)+","+x+"-last");
		Instances testData = Filter.useFilter(data, rm);
		System.out.println("testData size " + testData.numInstances());
		



		 FilteredClassifier fc = new FilteredClassifier();
		fc.setClassifier(new RandomForest());
		 fc.setFilter(rm);
		
		 String[] options = weka.core.Utils.splitOptions("-I 300 -K "+numFeatures+" -S "+seedNumber);
			fc.setOptions(options);
	//	fc.buildClassifier(data);
	    Evaluation eval_mal = new Evaluation(data);		
		
		System.out.println("Number of instances: " + data.numInstances()+" and number of authors: " + data.numClasses());
		
		
		String[] options1 = weka.core.Utils.splitOptions("-I 300 -K "+numFeatures+" -S "+seedNumber);
		cls.setOptions(options);
	//	cls.buildClassifier(data);
		
		

		Evaluation eval=null;
		

		if(endRelax==1)
		 eval = new Evaluation(data);
		else
		 eval= new RelaxedEvaluation(data, relaxPar);
		
		
				eval.crossValidateModel(cls, data,foldNumber , new Random(seedNumber));
			
				
				

				
			 
			     // generate curve
			     ThresholdCurve tc = new ThresholdCurve();
			     int classIndex = 0;
			     Instances result = tc.getCurve(eval.predictions(), classIndex);
			     System.out.println(tc.getROCArea(result));
				
				
		System.out.println("Relaxed by, "+relaxPar+", seedNo,"+seedNumber+", files,"+numberFiles+", authors,"+data.numClasses());
		Util.writeFile("Relaxed by, "+relaxPar+", seedNo,"+seedNumber+", files,"+numberFiles+", authors,"+data.numClasses(),
				fileName, true);

	     accuracy=eval.pctCorrect();
	     total =total+accuracy;
	     average = total/seedNumber;
			}	

				  System.out.println("total is "+total);
				  System.out.println("avg is "+average);
				  System.out.println("accuracy is "+accuracy);

		System.out.println("\nThe average accuracy with "+numberFiles+"files is "+average+"\n");	
	     Util.writeFile("\nThe average accuracy with "+numberFiles+"files is "+average+", relaxed by, "+relaxPar+", \n",
	    		 fileName, true);
	     
			  }
			  

	     }}
		}
	}
	
}