PREDICT CATEGORIES
Classification
A classification model predicts a category such as yes/no, spam/not spam, or one of several product types. Some models return probabilities; a separate decision rule turns those estimates into labels.
Check more than accuracy
Accuracy can hide poor performance when one class is rare. Depending on the cost of mistakes, examine precision, recall, a confusion matrix and probability calibration. Choose a threshold using the task’s real trade-offs, and assess the final model on data that was not used to fit it.
WORKED EXAMPLES
Classification guides
Each article states what its model predicts and how to evaluate it.