如何在 Java Weka API 中使用类不平衡技术 (SMOTE)？

Question

我正在尝试使用 Java Weka API 构建分类模型。我的训练数据集存在类别不平衡问题。出于这个原因，我想使用像 SMOTE 这样的类不平衡技术来减少类不平衡问题。源代码如下：package classification;import java.util.Random;import weka.classifiers.Classifier;import weka.classifiers.bayes.NaiveBayesMultinomial;import weka.core.Instance;import weka.core.Instances;import weka.core.converters.ConverterUtils.DataSource;import weka.filters.Filter;import weka.filters.unsupervised.attribute.StringToWordVector;public class questStackoverflow {public static void main(String agrs[]) throws Exception{String fileRootPath = "../file.arff"; //Dataset    Instances strdata = DataSource.read(fileRootPath); //Load Dataset    StringToWordVector filter = new StringToWordVector(10000);    filter.setInputFormat(strdata);    String[] options = { "-W", "10000", "-L", "-M", "1",            "-stemmer", "weka.core.stemmers.IteratedLovinsStemmer",             "-stopwords-handler", "weka.core.stopwords.Rainbow",             "-tokenizer", "weka.core.tokenizers.AlphabeticTokenizer"             };    filter.setOptions(options);    filter.setIDFTransform(true);    Instances data = Filter.useFilter(strdata,filter); //Apply filter    data.setClassIndex(0); //set class index            double recall=0.0;    double precision=0.0;    double fmeasure=0.0;    double tp, fp, fn, tn;    Classifier classifier = null;    classifier = new NaiveBayesMultinomial(); //classifer                }            }           }我的代码在没有类不平衡技术的情况下运行良好。但是，我需要使用类不平衡技术来缓解类不平衡问题。但是，我不知道如何在 Java Weka API 中使用它。

Answer 1

您可以在代码中添加以下代码行：

weka.filters.supervised.instance.SMOTE

SMOTE smote=new SMOTE();

smote.setInputFormat(trains);       

Instances Trains_smote= Filter.useFilter(trains, smote);

您的代码如下。

package classification;

import java.util.Random;

import weka.classifiers.Classifier;

import weka.classifiers.bayes.NaiveBayesMultinomial;

import weka.core.Instance;

import weka.core.Instances;

import weka.core.converters.ConverterUtils.DataSource;

import weka.filters.Filter;

import weka.filters.unsupervised.attribute.StringToWordVector;

weka.filters.supervised.instance.SMOTE

public class questStackoverflow {

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

String fileRootPath = "../file.arff"; //Dataset

Instances strdata = DataSource.read(fileRootPath); //Load Dataset

StringToWordVector filter = new StringToWordVector(10000);

filter.setInputFormat(strdata);

String[] options = { "-W", "10000", "-L", "-M", "1",

        "-stemmer", "weka.core.stemmers.IteratedLovinsStemmer", 

        "-stopwords-handler", "weka.core.stopwords.Rainbow", 

        "-tokenizer", "weka.core.tokenizers.AlphabeticTokenizer" 

        };

filter.setOptions(options);

filter.setIDFTransform(true);

Instances data = Filter.useFilter(strdata,filter); //Apply filter

data.setClassIndex(0); //set class index        

double recall=0.0;

double precision=0.0;

double fmeasure=0.0;

double tp, fp, fn, tn;

Classifier classifier = null;

classifier = new NaiveBayesMultinomial(); //classifer

int folds = 10;         

Random random = new Random(1);

data.randomize(random);

data.stratify(folds);

tp = fp = fn = tn = 0;

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

   Instances trains = data.trainCV(folds, i,random); //training dataset

   Instances tests = data.testCV(folds, i); //testing dataset

   SMOTE smote=new SMOTE();

   smote.setInputFormat(trains);        

   Instances Trains_smote = Filter.useFilter(trains, smote);

    classifier.buildClassifier(Trains_smote);    //build classifier           

    for (int j = 0; j < tests.numInstances(); j++) {    

       Instance instance = tests.instance(j);

       double classValue = instance.classValue();                   

       double result = classifier.classifyInstance(instance);

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

                tp++;

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

                fp++;

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

                fn++;

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

                tn++;

            }

        }   

    }

    if (tn + fn > 0)

        precision = tn / (tn + fn);

    if (tn + fp > 0)

        recall = tn / (tn + fp);

    if (precision + recall > 0)

        fmeasure = 2 * precision * recall / (precision + recall);

    System.out.println("Precision: " + precision);

    System.out.println("Recall: " + recall);

    System.out.println("Fmeasure: " + fmeasure);

}

}