# Machine Learning functions

**URL:** <https://discuss.elastic.co/t/machine-learning-functions/129990>\
**Category:** Elasticsearch\
**Tags:** elastic-stack-machine-learning\
**Created:** [April 30, 2018, 7:28am UTC](https://discuss.elastic.co/t/machine-learning-functions/129990 "2018-04-30T07:28:38Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![mkemalm](https://avatars.discourse-cdn.com/v4/letter/m/a3d4f5/32.png) [@mkemalm](https://discuss.elastic.co/u/mkemalm)\
**Post date:** [April 30, 2018, 7:28am UTC](https://discuss.elastic.co/t/machine-learning-functions/129990/1 "2018-04-30T07:28:38Z")

</div>

Hi,

in ML jobs functions are categorized in three groups; function itself, low function and high function. For example sum, low sum, high sum. ELK documents explain those functions as;

function itself detects anomalies in a bucket  
low detects anomalies in a usual low bucket (when sth begins to increase)  
high detects anomalies in a usual high bucket(when sth begins to decrease)

What is the point of using low or high instead of common anomaly detection function?

---

<div class="post-metadata">

**Author:** ![polyfractal](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/polyfractal/32/48162_2.png) [@polyfractal](https://discuss.elastic.co/u/polyfractal)\
**Post date:** [April 30, 2018, 4:22pm UTC](https://discuss.elastic.co/t/machine-learning-functions/129990/2 "2018-04-30T16:22:04Z")

</div>

There are some situations where you don't care about the value moving in one direction, but care if it goes the other way. So for `high_sum`, you only care if the value is anomalously high. If it's anomalously low that's ok. For example, if your websites latency is normally `100ms`, you may not care if it dips to `20ms`, but you certainly care if it spikes to `1000ms`. So a "high" function is what you're looking for.

Ditto for "low" functions, except the other direction.

The function itself is used if you care about _any_ change in the value, going high or low. It's basically high + low together in one function.

---

<div class="post-metadata">

**Author:** ![system](https://us1.discourse-cdn.com/elastic/original/3X/1/a/1ac57faf039f6b580b3f104ef42a2a89e41014de.png) [@system](https://discuss.elastic.co/u/system)\
**Post date:** [May 28, 2018, 4:22pm UTC](https://discuss.elastic.co/t/machine-learning-functions/129990/3 "2018-05-28T16:22:06Z")

</div>

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.