opendistro-for-elasticsearch / anomaly-detection

A machine learning plugin in Open Distro for real time anomaly detection on streaming data.
https://opendistro.github.io/for-elasticsearch-docs/docs/ad/
Apache License 2.0
78 stars 36 forks source link

Introduction of categorical features #404

Open Invadazoid opened 3 years ago

Invadazoid commented 3 years ago

Is your feature request related to a problem? Please describe. By default, the maximum number of features allowed per detector is 5. While that can be increased by dynamically adjusting the limit at the expense of precision/recall, is there any way to enable adding categorical features, beyond the current default requirement of each feature being a one-dimensional number?

Describe the solution you'd like Currently, only one-dimensional numbers can be added in a detector as a feature. It would be nice to enable adding arrays as features, or categorical features.

Describe alternatives you've considered Is this an inherent limitation of the Random Cut Forest implementation, or can this modification be achieved via tweaking a setting? If so, please add the corresponding setting in the documentation, or mention here as response.

kaituo commented 3 years ago

The feature is in our plan. This is not released yet. Will plan to support 2 categorical fields.