# Choosing a Classification probability threshold when using nested cross-validation

**URL:** <https://discourse.datamethods.org/t/choosing-a-classification-probability-threshold-when-using-nested-cross-validation/2802>\
**Category:** modeling strategy\
**Created:** [December 15, 2019, 4:16pm UTC](https://discourse.datamethods.org/t/choosing-a-classification-probability-threshold-when-using-nested-cross-validation/2802 "2019-12-15T16:16:49Z")\
**Posts on this page:** 1\
**Showing post:** 9

<div class="post-metadata">

**Author:** ![ljubomir](https://discourse.datamethods.org/user_avatar/discourse.datamethods.org/ljubomir/32/388_2.png) [@ljubomir](https://discourse.datamethods.org/u/ljubomir)\
**Post date:** [December 19, 2019, 1:34am UTC](https://discourse.datamethods.org/t/choosing-a-classification-probability-threshold-when-using-nested-cross-validation/2802/9 "2019-12-19T01:34:48Z")

</div>

Noted. But what to do if product requirements dictate sensitivity and specificity (a common scenario)? If you optimize something like Brier score or calibration instead, test may not meet the requirements even if it is theoretically optimal

---

_[View the full topic](https://discourse.datamethods.org/t/choosing-a-classification-probability-threshold-when-using-nested-cross-validation/2802)._