Weka各类分类器的使用（Java）

1.配置

MyEclipse2013+Weka3.6+libsvm3.18+Jdk1.7+Win8.1

2.小Tips

1). Java使用Weka

实现：

将安装文件夹里的weka.jar加入项目编译路径中

2). CSV文件可以转换成Arff文件

实现：

运行Weka的Explorer界面，打开csv文件，保存为arff文件。

注意：

如果有训练集和测试集，将训练集的Arff文件的标签头复制到测试集的Arff文件！

3). Java通过Weka使用LibSVM

实现：

将LibSVM文件夹里的libsvm.jar加入项目编译路径中

3.示例

public static void main(String[] args) {

       try {

          

           Classifier classifier1;

           Classifier classifier2;

           Classifier classifier3;

           Classifier classifier4;

 

          

           File inputFile = new File(

                  "C:\\Users\\zhangzhizhi\\Documents\\Everyone\\张志智\\总结积累\\Weka\\change_train.arff");// 训练语料文件

           ArffLoader atf = new ArffLoader();

           atf.setFile(inputFile);

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

          

           inputFile = new File(

                  "C:\\Users\\zhangzhizhi\\Documents\\Everyone\\张志智\\总结积累\\Weka\\change_test.arff");// 测试语料文件

           atf.setFile(inputFile);

           Instances instancesTest = atf.getDataSet(); // 读入测试文件

 

          

           instancesTest.setClassIndex(0);

           instancesTrain.setClassIndex(0);

 

          

           // 朴素贝叶斯算法

           classifier1 = (Classifier) Class.forName(

                  "weka.classifiers.bayes.NaiveBayes").newInstance();

           // 决策树

           classifier2 = (Classifier) Class.forName(

                  "weka.classifiers.trees.J48").newInstance();

           // Zero

           classifier3 = (Classifier) Class.forName(

                  "weka.classifiers.rules.ZeroR").newInstance();

           // LibSVM

           classifier4 = (Classifier) Class.forName(

                  "weka.classifiers.functions.LibSVM").newInstance();

 

          

           classifier4.buildClassifier(instancesTrain);

           classifier1.buildClassifier(instancesTrain);

           classifier2.buildClassifier(instancesTrain);

           classifier3.buildClassifier(instancesTrain);

          

          

          

           Evaluation eval = new Evaluation(instancesTrain);

          

           eval.evaluateModel(classifier4, instancesTest);

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

           eval.evaluateModel(classifier1, instancesTest);

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

           eval.evaluateModel(classifier2, instancesTest);

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

           eval.evaluateModel(classifier3, instancesTest);

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

          

       } catch (Exception e) {

           e.printStackTrace();

       }

    }

如果只有训练集，采用十交叉验证的方法，将上面的第5步和第6步更改为如下代码：

          

           Evaluation eval = new Evaluation(instancesTrain);

           eval.crossValidateModel(classifier4, instancesTrain, 10, new Random(1));

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

           eval.crossValidateModel(classifier1, instancesTrain, 10, new Random(1));

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

           eval.crossValidateModel(classifier2, instancesTrain, 10, new Random(1));

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

           eval.crossValidateModel(classifier3, instancesTrain, 10, new Random(1));

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

如果需要保存和加载分类器模型参数，在第5步和第6步之间加入如下代码：

          

           SerializationHelper.write("LibSVM.model", classifier4);

           SerializationHelper.write("NaiveBayes.model", classifier1);

           SerializationHelper.write("J48.model", classifier2);

           SerializationHelper.write("ZeroR.model", classifier3);

          

          

           Classifier classifier8 = (Classifier) weka.core.SerializationHelper.read("LibSVM.model");

           Classifier classifier5 = (Classifier) weka.core.SerializationHelper.read("NaiveBayes.model");

           Classifier classifier6 = (Classifier) weka.core.SerializationHelper.read("J48.model");

           Classifier classifier7 = (Classifier) weka.core.SerializationHelper.read("ZeroR.model");