OpenSearch/docs
Benjamin Trent afd90647c9
[ML] Adds feature importance to option to inference processor (#52218) (#52666)
This adds machine learning model feature importance calculations to the inference processor.

The new flag in the configuration matches the analytics parameter name: `num_top_feature_importance_values`
Example:
```
"inference": {
   "field_mappings": {},
   "model_id": "my_model",
   "inference_config": {
      "regression": {
         "num_top_feature_importance_values": 3
      }
   }
}
```

This will write to the document as follows:
```
"inference" : {
   "feature_importance" : {
      "FlightTimeMin" : -76.90955548511226,
      "FlightDelayType" : 114.13514762158526,
      "DistanceMiles" : 13.731580450792187
   },
   "predicted_value" : 108.33165831875137,
   "model_id" : "my_model"
}
```

This is done through calculating the [SHAP values](https://arxiv.org/abs/1802.03888).

It requires that models have populated `number_samples` for each tree node. This is not available to models that were created before 7.7.

Additionally, if the inference config is requesting feature_importance, and not all nodes have been upgraded yet, it will not allow the pipeline to be created. This is to safe-guard in a mixed-version environment where only some ingest nodes have been upgraded.

NOTE: the algorithm is a Java port of the one laid out in ml-cpp: https://github.com/elastic/ml-cpp/blob/master/lib/maths/CTreeShapFeatureImportance.cc

usability blocked by: https://github.com/elastic/ml-cpp/pull/991
2020-02-21 18:42:31 -05:00
..
community-clients [DOCS] Add Elixir Bulk Processor to community clients (#50630) 2020-01-15 16:08:56 -05:00
groovy-api Make sure to use the type _doc in the REST documentation. (#34662) 2018-10-22 11:54:04 -07:00
java-api [DOCS] Remove unneeded redirects (#50510) 2020-01-06 09:11:48 -06:00
java-rest [ML][Inference] don't return inflated definition when storing trained models (#52573) (#52580) 2020-02-20 19:47:29 -05:00
painless Scripting: Add char position of script errors (#51069) (#51266) 2020-01-21 13:45:59 -07:00
perl [DOCS] Various spelling corrections (#37046) 2019-01-07 14:44:12 +01:00
plugins [Docs]: Fix typo 'Got' -> 'Go' (#52603) 2020-02-21 10:25:13 +01:00
python Update version numbers in Elasticsearch-Py docs (#40355) 2019-04-02 12:16:24 -04:00
reference [ML] Adds feature importance to option to inference processor (#52218) (#52666) 2020-02-21 18:42:31 -05:00
resiliency [DOCS] Change http://elastic.co -> https (#48479) (#51812) 2020-02-03 09:50:11 -05:00
ruby [DOCS] Various spelling corrections (#37046) 2019-01-07 14:44:12 +01:00
src/test Test: match the \ilm/ and \slm/ test name too (#51811) (#51919) 2020-02-05 09:56:53 +00:00
README.asciidoc [DOCS] Clarify backport policy for important technical corrections. (#49131) 2019-11-20 10:57:38 -08:00
Versions.asciidoc [DOCS] Include docs on permanently unreleased branches only (#51743) 2020-02-11 11:24:13 -05:00
build.gradle [7.x] Permit EQL feature flag in release builds (#52201) (#52214) 2020-02-11 11:49:49 -05:00

README.asciidoc

The Elasticsearch docs are in AsciiDoc format and can be built using the
Elasticsearch documentation build process.

See: https://github.com/elastic/docs

=== Backporting doc fixes

* Doc changes should generally be made against master and backported through to the current version
  (as applicable).  
  
* Changes can also be backported to the maintenance version of the previous major version. 
  This is typically reserved for technical corrections, as it can require resolving more complex
  merge conflicts, fixing test failures, and figuring out where to apply the change.

* Avoid backporting to out-of-maintenance versions. 
  Docs follow the same policy as code and fixes are not ordinarily merged to
  versions that are out of maintenance.
  
* Do not backport doc changes to https://www.elastic.co/support/eol[EOL versions]. 

=== Snippet testing

Snippets marked with `[source,console]` are automatically annotated with
"VIEW IN CONSOLE" and "COPY AS CURL" in the documentation and are automatically
tested by the command `./gradlew -pdocs check`. To test just the docs from a
single page, use e.g. `./gradlew -ddocs integTestRunner --tests "*rollover*"`.

By default each `[source,console]` snippet runs as its own isolated test. You
can manipulate the test execution in the following ways:

* `// TEST`: Explicitly marks a snippet as a test. Snippets marked this way
are tests even if they don't have `[source,console]` but usually `// TEST` is
used for its modifiers:
  * `// TEST[s/foo/bar/]`: Replace `foo` with `bar` in the generated test. This
  should be used sparingly because it makes the snippet "lie". Sometimes,
  though, you can use it to make the snippet more clear. Keep in mind that
  if there are multiple substitutions then they are applied in the order that
  they are defined.
  * `// TEST[catch:foo]`: Used to expect errors in the requests. Replace `foo`
  with `request` to expect a 400 error, for example. If the snippet contains
  multiple requests then only the last request will expect the error.
  * `// TEST[continued]`: Continue the test started in the last snippet. Between
  tests the nodes are cleaned: indexes are removed, etc. This prevents that
  from happening between snippets because the two snippets are a single test.
  This is most useful when you have text and snippets that work together to
  tell the story of some use case because it merges the snippets (and thus the
  use case) into one big test.
      * You can't use `// TEST[continued]` immediately after `// TESTSETUP` or
      `// TEARDOWN`.
  * `// TEST[skip:reason]`: Skip this test. Replace `reason` with the actual
  reason to skip the test. Snippets without `// TEST` or `// CONSOLE` aren't
  considered tests anyway but this is useful for explicitly documenting the
  reason why the test shouldn't be run.
  * `// TEST[setup:name]`: Run some setup code before running the snippet. This
  is useful for creating and populating indexes used in the snippet. The setup
  code is defined in `docs/build.gradle`. See `// TESTSETUP` below for a
  similar feature.
  * `// TEST[warning:some warning]`: Expect the response to include a `Warning`
  header. If the response doesn't include a `Warning` header with the exact
  text then the test fails. If the response includes `Warning` headers that
  aren't expected then the test fails.
* `[source,console-result]`: Matches this snippet against the body of the
  response of the last test. If the response is JSON then order is ignored. If
  you add `// TEST[continued]` to the snippet after `[source,console-result]`
  it will continue in the same test, allowing you to interleave requests with
  responses to check.
* `// TESTRESPONSE`: Explicitly marks a snippet as a test response even without
  `[source,console-result]`. Similarly to `// TEST` this is mostly used for
  its modifiers.
  * You can't use `[source,console-result]` immediately after `// TESTSETUP`.
  Instead, consider using `// TEST[continued]` or rearrange your snippets.

  NOTE: Previously we only used `// TESTRESPONSE` instead of
  `[source,console-result]` so you'll see that a lot in older branches but we
  prefer `[source,console-result]` now.

  * `// TESTRESPONSE[s/foo/bar/]`: Substitutions. See `// TEST[s/foo/bar]` for
  how it works. These are much more common than `// TEST[s/foo/bar]` because
  they are useful for eliding portions of the response that are not pertinent
  to the documentation.
    * One interesting difference here is that you often want to match against
    the response from Elasticsearch. To do that you can reference the "body" of
    the response like this: `// TESTRESPONSE[s/"took": 25/"took": $body.took/]`.
    Note the `$body` string. This says "I don't expect that 25 number in the
    response, just match against what is in the response." Instead of writing
    the path into the response after `$body` you can write `$_path` which
    "figures out" the path. This is especially useful for making sweeping
    assertions like "I made up all the numbers in this example, don't compare
    them" which looks like `// TESTRESPONSE[s/\d+/$body.$_path/]`.
  * `// TESTRESPONSE[non_json]`: Add substitutions for testing responses in a
  format other than JSON. Use this after all other substitutions so it doesn't
  make other substitutions difficult.
  * `// TESTRESPONSE[skip:reason]`: Skip the assertions specified by this
  response.
* `// TESTSETUP`: Marks this snippet as the "setup" for all other snippets in
  this file. This is a somewhat natural way of structuring documentation. You
  say "this is the data we use to explain this feature" then you add the
  snippet that you mark `// TESTSETUP` and then every snippet will turn into
  a test that runs the setup snippet first. See the "painless" docs for a file
  that puts this to good use. This is fairly similar to `// TEST[setup:name]`
  but rather than the setup defined in `docs/build.gradle` the setup is defined
  right in the documentation file. In general, we should prefer `// TESTSETUP`
  over `// TEST[setup:name]` because it makes it more clear what steps have to
  be taken before the examples will work. Tip: `// TESTSETUP` can only be used
  on the first snippet of a document.
* `// TEARDOWN`: Ends and cleans up a test series started with `// TESTSETUP` or
  `// TEST[setup:name]`. You can use `// TEARDOWN` to set up multiple tests in
  the same file.
* `// NOTCONSOLE`: Marks this snippet as neither `// CONSOLE` nor
  `// TESTRESPONSE`, excluding it from the list of unconverted snippets. We
  should only use this for snippets that *are* JSON but are *not* responses or
  requests.

In addition to the standard CONSOLE syntax these snippets can contain blocks
of yaml surrounded by markers like this:

```
startyaml
  - compare_analyzers: {index: thai_example, first: thai, second: rebuilt_thai}
endyaml
```

This allows slightly more expressive testing of the snippets. Since that syntax
is not supported by `[source,console]` the usual way to incorporate it is with a
`// TEST[s//]` marker like this:

```
// TEST[s/\n$/\nstartyaml\n  - compare_analyzers: {index: thai_example, first: thai, second: rebuilt_thai}\nendyaml\n/]
```

Any place you can use json you can use elements like `$body.path.to.thing`
which is replaced on the fly with the contents of the thing at `path.to.thing`
in the last response.