通过weka.jar包来进行数据预处理

来源：互联网发布：js保留三位小数编辑：程序博客网时间：2024/05/01 07:41

打开eclipse ,在对应的工程下右击，选择Build Path ->选择Configure Build Path ->选择Libraries ->点击Add External JARs ->然后到你的jar包所在路径选择它。即可。

一、特征选择

[java] view plain copy
 print?
package learning;  
  
import weka.attributeSelection.ASEvaluation;  
import weka.attributeSelection.InfoGainAttributeEval;  
import weka.attributeSelection.Ranker;  
import weka.core.Instances;  
import weka.core.converters.ConverterUtils.DataSink;  
import weka.core.converters.ConverterUtils.DataSource;  
import weka.filters.Filter;  
import weka.filters.supervised.attribute.AttributeSelection;  
  
  
/**feature selection via weka 
 *  
 * @author wenbaoli 
 * 
 */  
public class featureSelect {  
  
    /** 
     *  
     * @param arg 
     */  
    public static void main(String[] arg){  
          
        try {  
  
            System.out.println("++++++++++++Example3:Feature Selection Via Weka.+++++++++");  
              
            System.out.println("Step1:load data...");  
            String fn = "E:/weka/data/iris.arff";  
            DataSource source = new DataSource(fn);  
            Instances instances = source.getDataSet();  
              
            System.out.println("Step2:feature selction...");  
            featureSelect fs = new featureSelect();  
              
            int k = 2;  
            AttributeSelection as = new AttributeSelection();  
              
            Ranker rank = new Ranker();  
            rank.setThreshold(0.0);  
            rank.setNumToSelect(k);  
              
            ASEvaluation ae = new InfoGainAttributeEval();  
          
            as.setEvaluator(ae);  
            as.setSearch(rank);  
            as.setInputFormat(instances);  
            Instances reductData = Filter.useFilter(instances, as);  
              
            System.out.println("Step3:保存规约后的数据到新文件...");  
            DataSink.write("E:/weka/data/iris_reducted.arff", reductData);  
            System.out.println("Finished...");  
              
              
        } catch (Exception e) {  
            e.printStackTrace();  
        }     
    }  
      
}  

二、缺失值处理

[java] view plain copy
 print?
package learning;  
  
import weka.core.Instances;  
import weka.core.converters.ConverterUtils.DataSink;  
import weka.core.converters.ConverterUtils.DataSource;  
  
  
/**Missing value Handling via weka 
 *  
 * @author wenbaoli 
 * 
 */  
public class missingHandle {  
  
    /** 
     *  
     * @param arg 
     */  
    public static void main(String[] arg) {  
          
        try {  
            System.out.println("+++++++++++++Example 2 :Missing Value Handling.++++++++++++++");  
              
            System.out.println("Step1:load data...");  
              
            String fn = "E:weka/data/labor.arff";  
              
            DataSource source = new DataSource(fn);  
              
            Instances instances = source.getDataSet();  
            int dim = instances.numAttributes();  
            int num = instances.numInstances();  
              
            System.out.println("Step2:缺失值处理...");  
            double[] meanV = new double[dim];  
            for (int i = 0; i < meanV.length; i++) {  
                meanV[i] = 0;  
                int count = 0;  
                for (int j = 0; j < num; j++) {  
                    if(!instances.instance(j).isMissing(i)){  
                        meanV[i] += instances.instance(j).value(i);  
                        count++;  
                    }  
                }  
                meanV[i] = meanV[i]/count;  
                System.out.println(meanV[i]);  
            }  
              
              
            for (int i = 0; i < meanV.length; i++) {  
                meanV[i] = 0;  
                int count = 0;  
                for (int j = 0; j < num; j++) {  
                    if(instances.instance(j).isMissing(i)){  
                        instances.instance(j).setValue(i, meanV[i]);  
                    }  
                }  
                  
                  
            }  
              
            System.out.println("Step3:保存数据到新文件...");  
              
            DataSink.write("E:weka/data/labor_missingValueHandled.arff", instances);  
            System.out.println("Finished.");  
        } catch (Exception e) {  
            e.printStackTrace();  
        }  
          
          
    }  
}  

三、归一化处理

[java] view plain copy
 print?
package learning;  
  
  
  
import weka.core.Attribute;  
import weka.core.Instance;  
import weka.core.Instances;  
import weka.core.converters.ConverterUtils.DataSink;  
import weka.core.converters.ConverterUtils.DataSource;  
import weka.filters.Filter;  
import weka.filters.unsupervised.attribute.Normalize;  
  
  
/**normalize data via weka 
 *  
 * @author wenbaoli 
 * 
 */  
public class normalizeTest {  
  
    /** 
     *  
     * @param arg 
     */  
    public static void main(String[] arg) {  
          
          
        String file = "cpu.arff";  
        String file_norm = "norm_" + file;  
        //对数据进行归一化  
        try {  
        System.out.println("+++++++++++++Example 1 : Normalize Data via weka.+++++++++");  
          
        System.out.println("Step1:读取数据...");  
        DataSource source = new DataSource("E:/Weka/data/" + file);  
        Instances instances = source.getDataSet();  
          
        System.out.println("Step2:原数据打印...");  
        System.out.println("---------------------------------");  
        int attributeNo = instances.numAttributes();  
        for (int i = 0; i < attributeNo; i++) {  
            Attribute attr = instances.attribute(i);  
            System.out.print(attr.name() + "\t");  
              
        }  
        System.out.println();  
          
        int instanceNo = instances.numInstances();  
        for (int i = 0; i < instanceNo; i++) {  
            Instance ins = instances.instance(i);  
            System.out.print(ins.toString() + "\t");  
            System.out.println();  
        }  
          
        System.out.println("Step3:归一化...");  
        Normalize norm = new Normalize();  
        norm.setInputFormat(instances);  
          
        //归一化关键步骤：  
        Instances newInstances = Filter.useFilter(instances, norm);  
          
        System.out.println("Step4:归一化之后的数据(打印)...");  
        System.out.println("---------------------------------");  
          
        //打印属性名  
        int numOfAttributes = newInstances.numAttributes();  
        for (int i = 0; i < numOfAttributes; i++) {  
            Attribute attribute = newInstances.attribute(i);  
            System.out.print(attribute.name() + "\t");  
              
        }  
        System.out.println();  
          
        //打印实例  
        int numOfInstance = newInstances.numInstances();  
        for (int i = 0; i < numOfInstance ; i++) {  
            Instance instance = newInstances.instance(i);  
            System.out.print(instance.toString() + "\t");  
            System.out.println();  
        }  
        //发现一个问题：这把标签label也给归一化了。。。。。。。。。。这样可以吗？？？？？？？  
          
        System.out.println("Step5:保存归一化的新数据到新文件...");  
        System.out.println("-----------------------");  
        DataSink.write("E:/Weka/data/" +file_norm, newInstances);  
        System.out.println("Congratulations.");  
        } catch (Exception e) {  
            e.printStackTrace();  
        }  
          
          
    }  
}  

0 0