liveBook · Manning

Precision

As explained in Fight Fraud with Machine Learning

Precision is the fraction of cases predicted as positive that are actually positive.

Precision measures the fraction of cases predicted as positive that are actually positive: true positives divided by true positives plus false positives. In fraud detection, it answers how many emails or transactions flagged as fraudulent really are fraudulent. section 1.6.3

A cautious detector that flags fraud only when highly confident tends to achieve high precision, but may miss fraudsters and produce lower recall. The right balance depends on operational cost: maximize precision when falsely flagging genuine activity is especially expensive. section 1.6.3

Fraud-alert review

If a system flags a group of emails or transactions as fraudulent, precision reports the share of those alerts confirmed as fraud. It distinguishes a cautious, trustworthy alert stream from one containing many genuine cases. section 1.6.3

Equation 1.1 Defining f-score as the harmonic mean of precision and recall
F-score = 2 * Precision * Recall / (Precision + Recall)

Compare precision with recall through the F-score, their harmonic mean, when selecting among fraud-detection systems. section 1.6.3

How Fight Fraud with Machine Learning covers it

How deep it goes
Defines the confusion-matrix ratio and explains threshold trade-offs with recall, but does not develop implementation procedures.
Examples
Uses fraudulent-email and fraudulent-transaction alerts to interpret correctly identified positive predictions.
What it assumes
Assumes familiarity with classification outcomes and confusion-matrix counts.
What you can do after
Evaluate fraud-detection systems and choose whether precision should outweigh recall when false alerts are costly.
How it connects
Connects precision to recall, F-score, confusion matrices, classification thresholds, and precision-recall curves.

Also explained in

Compare how 27 books explain precision →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Fight Fraud with Machine Learning. The section links point to the parts of the book it draws on.