AI Prompt Engineering and Key Concepts in Machine Learning and NLP Practice Test

A confusion matrix is a table that records, for each actual class, how many examples were predicted into each class. In NLP classification you use it to see, for example, how often a true positive is labeled correctly, how often negatives are mistaken for positives, and where classes get confused with one another. From this table you derive useful metrics, especially when data are imbalanced. Precision tells you, among all instances predicted as a given class, how many truly belong to that class. Recall tells you, among all actual instances of that class, how many the model correctly found. F1 combines precision and recall to give a balance between them. Accuracy measures overall correct predictions but can be misleading if one class dominates. AUROC assesses the model’s ability to rank positives higher than negatives across all thresholds, which remains informative when class frequencies differ.

Other options don’t fit because they describe different things: a chart of predicted label distribution doesn’t involve actual labels, a plot of training loss tracks learning over time, and a matrix of hyperparameters shows model settings rather than evaluation results.