elastic / elasticsearch

Free and Open, Distributed, RESTful Search Engine
https://www.elastic.co/products/elasticsearch
Other
68.57k stars 24.36k forks source link

Update Dataframe API docs reference SHAP feature importance #52268

Closed davidkyle closed 4 years ago

davidkyle commented 4 years ago

https://www.elastic.co/guide/en/elasticsearch/reference/current/ml-update-datafeed.html

Contains the paragraph:

Advanced configuration option. If set, feature importance for the top most important features will be computed. Importance is calculated using the SHAP (SHapley Additive exPlanations) method as described in Lundberg, S. M., & Lee, S.-I. A Unified Approach to Interpreting Model Predictions. In NeurIPS 2017..

This should be in the update data frame analytics docs but not here. This isn't an option for Datafeeds

elasticmachine commented 4 years ago

Pinging @elastic/es-docs (>docs)

elasticmachine commented 4 years ago

Pinging @elastic/ml-core (:ml)

droberts195 commented 4 years ago

I think this is caused by a mistake in the definition of thetag::indices[] tag in ml-shared.asciidoc. The definition of tag::num-top-feature-importance-values[] comes before end::indices[].

lcawl commented 4 years ago

Yes, I added a fix in https://github.com/elastic/elasticsearch/pull/52246 but will separate it out so I can do a quick merge.