weka java 交叉验证_交叉验证的Java weka实现，并保存和重载模型

package weka_test;

import java.io.File;

import java.io.IOException;

import weka.classifiers.Classifier;

import weka.classifiers.trees.J48;

import weka.core.Instance;

import weka.core.Instances;

import weka.core.converters.ArffLoader;

import weka.experiment.InstanceQuery;

import weka.classifiers.Evaluation;

import java.util.Random;

public class test {

/**

* oracleInput

* @return data

* @throws Exception

*/

public static Instances oracleInput() throws Exception{

InstanceQuery query = new InstanceQuery();

String sql = "SELECT to_char(z.cydate,'yyyy/mm') AS d,sum(z.bcmoney) as c FROM zybc z"

+ " WHERE to_char(z.cydate,'yyyy/mm') IS NOT NULL"

+ " GROUP BY to_char(z.cydate,'yyyy/mm') ORDER BY to_date(to_char(z.cydate,'yyyy/mm'),'yyyy/mm') ASC";

//System.out.println(sql);

query.setCustomPropsFile(new File("weka/weka_oracle.props"));

query.setDatabaseURL("jdbc:oracle:thin:@192.168.2.133:1521/XE");

query.setUsername("***");

query.setPassword("***");

query.setQuery(sql);

Instances data = query.retrieveInstances();

return data;

}

/**

* mysqlInput

* @return data

* @throws Exception

*/

public static Instances mysqlInput() throws Exception{

InstanceQuery query = new InstanceQuery();

String sql = "SELECT * FROM iris";

//System.out.println(sql);

query.setCustomPropsFile(new File("weka/weka_mysql.props"));

query.setDatabaseURL("jdbc:mysql://localhost:3306/test");

query.setUsername("***");

query.setPassword("***");

query.setQuery(sql);

Instances data = query.retrieveInstances();

return data;

}

/**

* @param args

* @throws Exception

*/

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

// TODO Auto-generated method stub

Classifier m_classifier = new J48();

/*File inputFile = new File("D://Program Files//Weka-3-7//data//iris.arff");//训练语料文件

ArffLoader atf = new ArffLoader();

atf.setFile(inputFile);

Instances instancesTrain = atf.getDataSet(); // 读入训练文件 */

Instances Train = mysqlInput();

Instances Test = mysqlInput();

Test.setClassIndex(4); //设置分类属性所在行号(第一行为0号)，instancesTest.numAttributes()可以取得属性总数

double sum = Test.numInstances(),//测试语料实例数

right = 0.0f;

Train.setClassIndex(4);

m_classifier.buildClassifier(Train); //训练

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

//2、利用模型进行预测

int a=0,b=0,c=0,d=0;//记录每个类别的个数，方便计算评价指标

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

double classification = m_classifier.classifyInstance(Train.instance(i));

double classValue = Train.instance(i).classValue();

if (classification == 0.0 && classValue == 0.0) {

a++;

} else if (classification == 0.0 && classValue == 1.0) {

b++;

} else if (classification == 1.0 && classValue == 0.0) {

c++;

} else if (classification == 1.0 && classValue == 1.0) {

d++;

}

}

// 3、得出预测效果评测指标

double precision = (double) a / (a + b);

double recall = (double) a / (a + c);

double fMeasure = 2 * precision * recall / (precision + recall);

System.out.println("precision\trecall\tF-Measure");

System.out.println((precision) + "\t\t"

+ (recall) + "\t"

+ (fMeasure));

for(int i = 0;i

{

if(m_classifier.classifyInstance(Test.instance(i))==Test.instance(i).classValue())//如果预测值和答案值相等(测试语料中的分类列提供的须为正确答案，结果才有意义)

{

right++;//正确值加1

}

}

String s=right+","+sum+",";

System.out.println("classification precision:"+s+(right/sum));

//直接调用Evaluation即可完成

Evaluation eval = null;

for (int i = 0; i < 10; i++) {

eval = new Evaluation(Train);

eval.crossValidateModel(m_classifier, Train, 10, new Random(i),

args);// 实现交叉验证模型

}

System.out.println(eval.toSummaryString());// 输出总结信息

System.out.println(eval.toClassDetailsString());// 输出分类详细信息

System.out.println(eval.toMatrixString());// 输出分类的混淆矩阵

}

}