[DOCS] Adds data frame analytics API and evaluate API resource documentation #43972
Conversation
szabosteve
added
:ml
|
Pinging @elastic/ml-core |
|
Pinging @elastic/es-docs |
| mertics: | ||
|
|
||
| `auc_roc` (Optional):: | ||
| (object) The AUC ROC score and optionally the curve. |
There was a problem hiding this comment.
- For each metric we should probably have another section where we explain each object.
- Also, I wonder if we should add links to wikipedia pages here and for the other metrics.
- Finally, let's also explain what auc_roc stands for: area under the curve of the receiver operating characteristic
There was a problem hiding this comment.
- I'll add these sections later as it takes some more time to produce them.
- I haven't seen such external references, but I'll ask if it is possible to add.
- The latest commit explains the abbreviation.
|
One last comment. In the description of the default behaviour of the method parameter, I'd change the sentence:
to
The technical term for what we do is to create an ensemble. The way we combine the results is not really hierarchical so I would just say combine and not be specific. We can provide more technical details of exactly how we combine the results, but it would probably be too much to include in the API docs. However, knowing that we use a normalised combination of individual method scores is useful. |
There was a problem hiding this comment.
Couple of minor points on model_memory_limit, but otherwise looks good @szabosteve.
|
@elasticmachine update branch |
|
merge conflict between base and head |
| [testenv="platinum"] | ||
| [[ml-evaluate-dfanalytics-resources]] | ||
| === {dfanalytics-cap} evaluation resources | ||
|
|
There was a problem hiding this comment.
This page is different than most of the other pages in the "Definitions" section, since it seems to be defining the input (request body) properties for the evaluate DF analytics API, rather than the output (response body) properties. In many other cases, the input and output is similar (i.e. input to create job matches output from get jobs so the "job resources" applies to both). That doesn't seem to be the case here, though.
I think we should either (a) extend the evaluation resources page to also describe the response objects, or (b) move the configuration objects into the API reference page and only cover the response objects in the resources page.
If I've explained this poorly or misunderstood the goal of this page, just let me know!
There was a problem hiding this comment.
The resources page contains both the request body and the response body parameters.
Some params are overlapping, for example auc_roc, precison, recall, confusion_matrix can be also the part of the request body as well as of the response body. tp, fp, tn, fn could be only the part of the response body. As far as I see, all the response objects are covered here.
| [[binary-sc-resources-properties]] | ||
| ===== {api-definitions-title} | ||
|
|
||
| `actual_field` (Required):: |
There was a problem hiding this comment.
Typically this information about which fields are required or optional appears in the API reference page. It's unusual to see it here.
There was a problem hiding this comment.
I removed the notes in fc3ad12.
…ntation (#43972) This PR adds the resource documentation of the data frame analytics APIs and the evaluate API to the ML API doc pool.
…ntation (#43972) This PR adds the resource documentation of the data frame analytics APIs and the evaluate API to the ML API doc pool.
|
Backported to 7.x and 7.3. |
This PR adds the resource documentation of the data frame analytics APIs and the evaluate API to the ML API doc pool.
NOTE: This PR must be merged after elastic/docs#968