In a confusion matrix, what does the term true positive represent ?

Concept: In a confusion matrix, every prediction on a binary classifier is placed into one of four cells based on two independent facts: the instance's ACTUAL class (positive or negative) and the class the model PREDICTED for it (positive or negative). Each cell's name states these two facts in order - whether the prediction was correct ("true") or wrong ("false"), followed by what was predicted ("positive" or "negative").

Application: Applying this naming rule to "true positive": "true" means the prediction matched the actual class, and "positive" means the model predicted the positive class. So a true positive is an instance whose actual class is positive and which the model also predicted as positive - an accurate identification of a positive case. Among the given options, that is the accurate prediction of positive cases.

Cross-check: The remaining three cells follow the same rule: an accurate prediction of negative cases is a true negative (actual negative, predicted negative); positive cases incorrectly predicted as negative form a false negative (actual positive, predicted negative); and negative cases incorrectly predicted as positive form a false positive (actual negative, predicted positive). Crossing "actual" against "predicted" this way for all four combinations confirms that "accurate prediction of positive cases" is the one that matches true positive.