weka属性选择使用

醉了～～～
package edu.dcy.weka;

import java.io.FileWriter;

import java.util.ArrayList;

import java.util.List;

import weka.classifiers.Classifier;

import weka.classifiers.Evaluation;

import weka.classifiers.misc.InputMappedClassifier;

import weka.core.Attribute;

import weka.core.Instance;

import weka.core.Instances;

import weka.core.SerializationHelper;

import weka.core.Utils;

import weka.core.converters.ConverterUtils.DataSource;

public class PredictModel {

    public static void main(String[] args) throws Exception {

        if(args.length!=3){

            System.err.println("Usage: <model file> <test file > < output dir>");

            System.exit(1);

        }

        // load data

        System.out.println("loading data.....");

        DataSource dstest = new DataSource(args[1]);

        Instances test = dstest.getDataSet();

        int attrnums=test.numAttributes();

        // lack of the class lable 'Y'

        if(attrnums==283){

            List<String> strname = new ArrayList<String>();

            strname.add("false");

            strname.add("true");

            Attribute attr = new Attribute("Y", strname);

            test.insertAttributeAt(attr, test.numAttributes());

        }

        test.setClassIndex(test.numAttributes() - 1);

        System.out.println("loading done.....");

        System.out.println("loading predict model.....");

    	// load classifier

		Object objs[]=SerializationHelper.readAll(args[0]);

		Classifier cls=(Classifier) objs[0];

		InputMappedClassifier mapper =new InputMappedClassifier();

		mapper.setClassifier((Classifier) objs[0]);

		mapper.setModelHeader((Instances) objs[1]);

		mapper.setTestStructure(test);

		cls=mapper;

        // output predictions

        System.out.println("execute ...");

        FileWriter fw=new FileWriter(args[2]);

        StringBuffer sb=new StringBuffer();

        sb.append("#,actual,predicted,error,p_false,p_true \n");

        for (int i = 0; i < test.numInstances(); i++) {

            Instance ins=test.instance(i);

            double pred = cls.classifyInstance(ins);

            double[] dist = cls.distributionForInstance(ins);

            sb.append(i+1).append(",")

            .append(ins.toString(test.classIndex()))

            .append(",")

            .append(test.classAttribute().value((int) pred))

            .append(",");

            if (pred != test.instance(i).classValue()) {

                sb.append("yes");

            } else {

                sb.append("no");

            }

            sb.append(",").append(Utils.arrayToString(dist)).append("\n");

//            System.out.println(sb.toString());

            fw.write(sb.toString());

            fw.flush();

            sb.delete(0, sb.length());

        }

        fw.close();

        System.out.println("finished,please check the outfile .....");

        // print the summary

        if(attrnums==284){

            Evaluation eval = new Evaluation(test);

            eval.evaluateModel(cls, test);

            System.out.println(eval.toClassDetailsString());

            System.out.println(eval.toSummaryString());

            System.out.println(eval.toMatrixString());

        }

    }

}
巴特西

weka属性选择使用

最新文章

热门文章
