Precision-recall curve

As explained in Grokking Statistics

A precision-recall curve plots a classifier's precision against its recall across decision thresholds.

A precision-recall curve plots a classifier’s precision against its recall across decision thresholds. Each threshold creates a different confusion matrix, so the curve shows how the model trades false alarms for missed positives. section 8.10

Use the curve to compare threshold choices:

• Raising the threshold can increase precision while decreasing recall. section 8.10
• Prefer a precision-recall curve when the classes are imbalanced, such as when non-spam messages greatly outnumber spam. section 8.11
• Use area under the curve as a single-number summary, rather than relying only on accuracy. section 8.11

Predicting toddler meltdowns

A logistic regression predicts whether a toddler will have a meltdown based on hours without a nap. Testing many thresholds produces the precision-recall curve: a stricter threshold catches fewer meltdowns but makes fewer false alarms, while a looser threshold catches more meltdowns at the cost of more false alarms. A larger dataset makes the curve more informative than a tiny sample. section 8.10

Figure 8.9 The precision-recall curve (PRC) for our toddler meltdown dataset

Inspect the precision-recall curve for the toddler meltdown dataset and connect its shape to threshold choices. section 8.10

How Grokking Statistics covers it

How deep it goes

Explains threshold tradeoffs, confusion matrices, dataset size, and AUC, but does not derive the curve mathematically.

Examples

Uses toddler meltdown prediction, nap duration, a 100-sample dataset, Figure 8.9, and scikit-learn curve code.

What it assumes

Assumes basic classification and precision-recall terminology while explaining threshold effects through confusion matrices.

What you can do after

Evaluate classifier thresholds with a precision-recall curve and summarize performance using AUC.

How it connects

Connects precision and recall with thresholds, confusion matrices, logistic regression, ROC curves, AUC, and cross-validation.

liveBookmanning.com

Read Grokking Statistics on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Machine Learning Algorithms in DepthA precision-recall curve plots precision against recall across classification thresholds.

Compare how 2 books explain precision-recall curve →

Related concepts

• area under the curvemeasured by · The area under a curve is the integral of its values over a specified domain and, in classification evaluation, commonly summarizes performance across decision thresholds.
• recalluses · Recall is the fraction of relevant, positive, or target outcomes that a system successfully identifies, retrieves, or captures.
• precisionuses · Precision is the proportion of items identified as positive, selected, or retrieved that satisfy the relevant correctness or relevance criterion.
• receiver operating characteristicalternative to · A receiver operating characteristic curve plots a binary classifier’s true-positive rate against its false-positive rate as the decision threshold varies.
• class imbalanceused by · Class imbalance is a condition in a classification dataset where classes have substantially unequal numbers of examples.
• fraud detectionmeasures · Fraud detection is the process of identifying transactions, activities, or other events suspected of involving intentional deception or unauthorized gain.

This explainer was generated by Manning's knowledge graph from the text of Grokking Statistics. The section links point to the parts of the book it draws on.