Question

我已经在svm中训练了smo模型，并使用SerializationHelper将其写入文件。

当我尝试读取模型进行分类时，会出现以下错误：

发现问题：java.io.StreamCorruptedException：无效的流标题：47656E65

以下是给出错误的行：

classifier = (Classifier) SerializationHelper.read(strModel);

编辑：

    Instances inst = Filter.useFilter(data, vectorizer);
    inst.setClassIndex(0);
    SMO svm = new SMO();
    RBFKernel kernal = new RBFKernel();
    svm.setKernel(kernal);
    svm.buildClassifier(filteredDate);
    SerializationHelper.write("SMO_samp.model", svm);

Answer 1

对于错误，this可能会有所帮助。

这是一个完全有效的例子：

import java.io.BufferedReader;
import java.io.FileReader;
import java.util.Random;
import weka.classifiers.Classifier;
import weka.classifiers.Evaluation;
import weka.classifiers.bayes.NaiveBayes;
import weka.classifiers.functions.LibSVM;
import weka.classifiers.functions.SMO;
import weka.core.Instances;
import weka.core.SerializationHelper;


public class CodeTest {

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

        BufferedReader breader = null;
        breader = new BufferedReader(new FileReader("iris.arff"));

        Instances train_data = new Instances (breader);
        breader.close();
        train_data.setClassIndex(train_data.numAttributes() -1);

        SMO smo = new SMO();
        smo.buildClassifier(train_data);

        SerializationHelper s = new SerializationHelper();
        s.write("SMO_samp.model", smo);
        Classifier c = new SMO();
        c = (Classifier) s.read("SMO_Samp.model");
        Evaluation evaluation = null;
        evaluation = new Evaluation (train_data);
        evaluation.crossValidateModel(c, train_data, 10, new Random(1));
        System.out.println(evaluation.toSummaryString());
    }
}

阅读已经训练过的SMO模型会产生StreamoutofException错误

1 个答案: