[Lens] ML predictive layers

[flash1293]

The Elastic stack ML integration provides functionality to predict a time series. Right now these ML jobs have to be configured in the ML app, separate from the often used Dashboard/Lens/Discover apps. Adding a separate "predictive" layer type in Lens would be a nice and intuitive integration of these capabilities in a prominent spot.

[elasticmachine]

Pinging @elastic/kibana-app (Team:KibanaApp)

[wylieconlon]

@flash1293 Are you referring to anomaly detection and forecasting as "predictive layers"? I think both of them fall into the category of solution-specific datasources. It is technically possible for a user to work around this if:

1. They have defined an ML index pattern, which is not usually done
2. They are able to filter to a specific model's results using KQL, for example result_type: model_plot instead of result_type: model_forecast.
3. They are able to select the date histogram + min/max aggregation to create a "bounded" prediction for each time range.

But I think that overall users would be happier if they could follow a simpler set of steps, for example:

1. Select from a dropdown of existing ML jobs
2. Select the type of data to load, for example "anomalies", "predicted trends" or "forecast"

[flash1293]

@wylieconlon I was focused on the forecasting, but it makes sense to broaden the scope.

[markov00]

In order to provide better transparency of priorities, issues that will not be prioritized within the next 24 months are being closed.

Tracking request in Lens general improvements ice box #184459