Question

Weka使用的概率平均值是多少？ distributionForInstanceAverage里面的代码是什么？

Answer 1

它只返回每个基本分类器的概率分布的均值（在投票中学习或在外部构建并由投票加载）。

首先对在Vote中学习的任何模型进行总结，然后加载任何模型，然后按元素方式将数组除以模型总数。

protected double[] distributionForInstanceAverage(Instance instance) throws Exception {

    //Init probs array with first classifier used within model (learnt or loaded)
    double[] probs = (m_Classifiers.length > 0) 
    ? getClassifier(0).distributionForInstance(instance)
        : m_preBuiltClassifiers.get(0).distributionForInstance(instance);

    //Add the distributions of any classifiers built within the Vote classifier to probs array
    for (int i = 1; i < m_Classifiers.length; i++) {
      double[] dist = getClassifier(i).distributionForInstance(instance);
      for (int j = 0; j < dist.length; j++) {
          probs[j] += dist[j];
      }
    }

    //Add the distributions of any classifiers built outside of the Vote classifier (loaded in) to the probs array
    int index = (m_Classifiers.length > 0) ? 0 : 1;
    for (int i = index; i < m_preBuiltClassifiers.size(); i++) {
      double[] dist = m_preBuiltClassifiers.get(i).distributionForInstance(instance);
      for (int j = 0; j < dist.length; j++) {
        probs[j] += dist[j];
      }
    }

    //Divide each probability by the total number of classifiers used in Vote (to get the mean)
    for (int j = 0; j < probs.length; j++) {
      probs[j] /= (double)(m_Classifiers.length + m_preBuiltClassifiers.size());
    }
    return probs;
  }

什么是平均概率？在“投票”部分内

1 个答案: