WEKA API和WEKA GUI为交叉验证提供了不同的输出

Question

我正在尝试为我的特定问题建立交叉验证过程。我将我的代码和WEKA GUI输出进行了比较，但是正确分类和错误分类实例的数量之间存在差异。我可以错过什么吗？

下面给出了我的代码（使用WEKA API版本3.8.3，我也尝试了3.8.2）：

private static void crossValidate(Instances instances, Classifier classifier, int k_value) throws Exception {
    int seed = 1;
    int folds = k_value;
    Random rand = new Random(seed);
    Instances randData = new Instances(instances);
    randData.randomize(rand);
    int correctlyClassified = 0;
    int incorrectlyClassified = 0;
    for(int n = 0; n < folds; n++) {
        Instances train = randData.trainCV(folds, n, rand);
        Instances test = randData.testCV(folds, n);
        classifier.buildClassifier(train);
        for (int t = 0; t < test.numInstances(); t++) {
               Instance  aTestInstance = test.instance(t);
               String actualClass = test.classAttribute().value((int)aTestInstance.classValue());
               String predictedClass = test.classAttribute().value((int)classifier.classifyInstance(aTestInstance));
               if(actualClass.equals(predictedClass)) correctlyClassified++;
               else incorrectlyClassified ++;
        }
    }
    System.out.println("Total True is:" + correctlyClassified + "/" + incorrectlyClassified);
}