An ML team is organizing the metrics to include in a classification model's evaluation report. What is the name of the most basic metric, used to evaluate a classification model, that represents the proportion of predictions that were classified correctly out of all predictions?
Choosing the basic classification metric (accuracy).
Precision
Precision is the proportion of items predicted as positive that were actually positive.
It measures the quality of positive predictions and has a different denominator from the proportion correct out of all predictions (accuracy), so it is incorrect.
Accuracy
Correct. Accuracy is the most basic metric that represents the proportion of predictions classified correctly out of all predictions. However, it can be misleading when the classes are heavily imbalanced, so it is used alongside other metrics.
Recall
Recall is the proportion of actual positives that were correctly found.
It measures how few cases are missed, not the overall proportion correct (accuracy), so it is incorrect.
F1 score
The F1 score is the harmonic mean of precision and recall.
It is a combined metric that expresses the balance of the two, not the most basic metric of overall proportion correct (accuracy), so it is incorrect.
Organize the basic evaluation metrics for classification models.
・Accuracy: the proportion of all predictions classified correctly. The correct answer here.
・Precision: the proportion of items predicted as positive that were actually positive.
・Recall: the proportion of actual positives that were correctly predicted as positive.
・F1 score: the harmonic mean of precision and recall, representing the balance of the two.
Accuracy is intuitive, but it tends to look high on heavily imbalanced data, so use it alongside other metrics.