What are the concepts of precision and recall in machine learning?
In: Mathematics
Precision: Of the ones classified as positive, what proportion were correct?
Recall: Of all the positives, what proportion did you classify as positive?
A lot of machine learning problems can be formulated as classification problems. Which one of several types (classes) are a set of objects? For simplicity, lets consider a binary classification problem (two classes). Let’s call the classes positive and negative.
We have a group of objects. Some are positive and some are negative. We then have a model which seeks to classify them.
Precision and recall are measures of accuracy for a classifier like our model.
Precision is, of the objects identified as being from the class of interest, how many actually are? What fraction of the objects the model identified as being from the positive class are, in fact, from the positive class?
Recall is, of the objects that are truly from the class of interest, what fraction did the model identify? Of the objects from the positive class, what fraction did the classifier label as positive?
Latest Answers