The likelihood that the predicted category in a k-Nearest Neighbors or Naïve Bayes model is correct is known as the ______ percentage.

• A. Accuracy
• B. Recall
• C. Confidence
• D. Precision

Answer: (C)

This question is about the probability that a model’s predicted category is correct for a given instance, i.e., the prediction’s confidence. In kNN, you can view confidence as the fraction of the k neighbors that vote for the predicted class; that proportion serves as an empirical probability that the prediction is correct for that instance. In Naive Bayes, the model provides posterior probabilities for each class given the features, and the predicted class is the one with the highest posterior—that probability is the model’s confidence in that prediction. This differs from accuracy, recall, and precision, which are dataset-wide metrics: accuracy is overall correctness, recall is the proportion of actual positives captured, and precision is the proportion of predicted positives that are truly positive. Confidence is a per-instance measure of how likely the prediction is to be correct, and it can be useful for deciding when to trust a prediction or to abstain if confidence is low.