Recall is defined as the number of true positive (TP) predictions, divided by the sum of the number of true positives and false negatives (TP+FN): $$\text{Recall}=\frac{tp}{tp+fn} \, $$ It is also referred to as the True Positive Rate (TPR) or Sensitivity. See: http://en.wikipedia.org/wiki/Precision_and_recall Recall is defined only for a specific class value, and should thus be labeled with the class value for which is was computed. Use the mean_weighted_recall for the weighted average over all class values.
WEKA's Evaluation.truePositiveRate(int classIndex):
/**
* Calculate the true positive rate with respect to a particular class.
* This is defined as
*
* correctly classified positives
* ------------------------------
* total positives
*
*
* @param classIndex the index of the class to consider as "positive"
* @return the true positive rate
*/
public double truePositiveRate(int classIndex) {
double correct = 0, total = 0;
for (int j = 0; j < m_NumClasses; j++) {
if (j == classIndex) {
correct += m_ConfusionMatrix[classIndex][j];
}
total += m_ConfusionMatrix[classIndex][j];
}
if (total == 0) {
return 0;
}
return correct / total;
}| Minimum value | 0 |
| Maximum value | 0 |
| Unit | |
| Optimization | Higher is better |