Confusion matrix

	Negative	Positive
Negative	True Negative (TN)	False Positive (FP)
Positive	False Negative (FN)	True Positive (TP)

Accuracy

Accuracy = \frac{T N + T P}{T N + F P + T P + F N} = \frac{True Results}{Total Results}

Represents the number of correctly classified data instances over total number of data instances

Not a good metric when data set is unbalanced where both negative and positive classes have different number of data instances

e.g. 90 TN, 10 FN, 0 TP and FP will still result in 90% accuracy despite 100% of positive being classified as negative

Also known as the positive predictive value

Precision = \frac{T P}{T P + F P} = \frac{True Positive}{Total Positives}

Precision should ideally be 1, when $T P = T P + F P$ where $F P = 0$ i.e. no false positives.

Also known as the sensitivity or true positive rate

Recall = \frac{T P}{T P + F N} = \frac{True Positive}{Total Actual Positives}

Recall should ideally be 1, when $T P = T P + F N$ where $F N = 0$ i.e. no false negatives.

In an ideal classifier, precision and recall should be one. Hence, we should find a metric that takes into account both precision and recall.

F1 Score = 2 * \frac{Precision * Recall}{Precision + Recall}

F1 score only becomes high when both precision and recall are high.

Also known as the harmonic mean of precision and recall.

Measures the ability to correctly identify negative instances. Also known as the true negative rate

Specificity = \frac{T N}{T N + F P} = \frac{True Negative}{Total Actual Negatives}