It is important to validate a classifier, that is, to measure its classification error
rate, before deciding to use it for an application. Consider an example of a classification
problem where a classifier has to predict, based on some inputs (the exact
inputs are not relevant here), whether a person is suffering from a particular disease
X or not. A positive prediction says that the person has the disease, and a
negative prediction says the person does not have the disease. (The terminology
of positive/negative prediction can be used for any binary classification problem,
not just disease classification.)