Hello, thank you for a good article! I try to understand the holiday effect for region SE. It seems like the underlaying holiday input table doesn't contain holiday for region SE ( select * from `bigquery-public-data`.`ml_datasets`.`holidays_and_events_for_forecasting`
where region = 'SE' gives no rows) but still it provides an effect in forecasts (for example new year) when comparing models with or without holiday SE. Did you experience the same?
Hello, thank you for a good article! I try to understand the holiday effect for region SE. It seems like the underlaying holiday input table doesn't contain holiday for region SE ( select * from `bigquery-public-data`.`ml_datasets`.`holidays_and_events_for_forecasting`
where region = 'SE' gives no rows) but still it provides an effect in forecasts (for example new year) when comparing models with or without holiday SE. Did you experience the same?
OMG, I didn't know that it did not exist. However, you can easily create custom holidays if you need to https://cloud.google.com/bigquery/docs/reference/standard-sql/bigqueryml-syntax-create-time-series#custom_holidays
I am just about to add "julafton" in my custom holiday expression! ;)
Hi, I'm new to GCP, is there any step by step instructions available to implement this cost anomaly detection in GCP?
Thank you for sharing this, this is awesome!
Quick question, how often would you recommend running this process?
I’m glad you like it. We run it every day, the cost is pretty small
Do you use the https://hub.getdbt.com/kristeligt-dagblad/dbt_ml/0.5.1/