Training a model is only half the job; you also need to know how good it is. Accuracy alone can be misleading, especially when one class is much rarer than the other, so this lesson introduces the metrics that tell you what a model is really getting right and wrong.
You'll start with the confusion matrix (true and false positives and negatives), then learn sensitivity and specificity, closely related to precision and recall. Finally you'll see how ROC curves and AUC summarize performance across every decision threshold, so you can compare models and pick a sensible cutoff.
Watch the videos in order, then try computing these metrics on a classifier of your own. When you're done, continue on to Data Splitting & Cross-validation.
How a confusion matrix shows what a classifier got right and wrong, and how it helps you decide which machine learning method to use.
Two key concepts for evaluating machine learning methods, and how they make it easier to choose between models.
How ROC graphs and AUC consolidate the information from many confusion matrices, and how to use them to pick a threshold and compare models.