I'm trying to create a forecast from an anomaly detection job in Kibana. It tracks the number of events in certain intervals, and we always have lower volume on certain holidays. I have applied a Calendar that I created to the job, and according to the documentation this will mean that the model / model bounds will not be skewed by the expected low-volume days.
My question though has to do with how this holiday calendar is applied to forecasts on the same job. When looking out at the forecast it doesn't account for these holidays and instead predicts a regular day's volume. And in the visualization there is not even annotation that it is a holiday, although the calendar applied included holidays for those future dates.
Is there some way to have the calendar applied to forecasts. I've seen this feature in other services, for example AWS - holidays featurization.
If it's not possible to have it apply to the prediction, is there at least some way to mark the graph in the metric viewer to visually highlight that the date represents a holiday?
Also, I just heard back from a colleague on the ML team.
He says the only way to display the calendar currently is to manually add an annotation (by dragging across in the Single Metric Viewer chart)
I just a short meeting the colleague I mentioned. He showed me how to add the annotation.
However, I know you want it to be automatic. So, I'll be adding an issue to the kibana repo (I'll post that here shortly, and then cross posting that issue in the ML enhancements repo.
That repo is private, but at least you'll be able to view activity in the Kibana issue.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.