ML 5.3Lesson outlineLesson 3 of 5 in Module 5
Precision–Recall Curves and Choosing a Threshold
Choose the operating threshold from the cost of each error or the capacity to act, and tune it without touching the test set.
Video in production.
Lesson outline: Precision–Recall Curves and Choosing a Threshold
In this lesson9 parts
- 01A curve that looks fine
- 02The training calls and five columns
- 03The default threshold
- 04Precision against recall, on ten clients
- 05Cost per threshold
- 06Two ways to read a threshold
- 07Tuning without the test set
- 08Code
- 09Recap, your challenge, and what comes next
Builds on
This lesson stands on its own. These go deeper into what it uses.
- ML 5.2: Thresholds, ROC and AUC
- 2.2
- 4.4; Stats 3.4 is taken at Stage 6
- after this video
- so the expected cost is computed from counts in the video (checked 2026-10-15)