What is a classifier?
A classifier is a model that performs classification: it takes an input and returns a category. The input might be a transaction record, an image, an email, or a window of sensor readings. The output is typically a score or probability for each possible class, and a decision rule picks the class, either the highest score or the one that clears a confidence threshold. Logistic regression, decision trees, support vector machines, naive Bayes, and neural networks are all classifiers once trained for a sorting task. A large language model given a prompt that asks it to label a document is acting as a classifier too, without any task-specific training.
Classifiers in practice
A managed service provider puts a classifier in front of its service desk mailbox. Each incoming message is scored for whether it is a support request, a sales inquiry, a vendor notification, or spam. Requests go straight to the ticketing system, and anything the model scores below the confidence threshold waits for a dispatcher. The classifier’s quality is tracked with a confusion matrix and with precision and recall per class, computed each month on a sample that a person has checked. When a new client with unusual email formats joins, the misclassification rate rises, and the model is retrained on the new examples. The monthly check is scheduled rather than left to chance, because classifier quality drifts quietly between evaluations.