Prediction Sync keeps your predictions synchronized by writing predictions to your target environment as new predictions are generated with the Keep Up To Date option.
What is Prediction Sync?
Prediction Sync allows you to specify the name of a prediction dataset so Qlik AutoML can automatically read your apply dataset, generate new predictions and write them back to your target environment on a scheduled basis. Prediction Sync requires that Keep Up To Date is enabled and AutoML will enable it if it's not already. Prediction Sync is available for all predictions EXCEPT those powered by CSV file upload. (.CSV files coming from buckets like Amazon S3 and MinIO are supported.)
How do I use Prediction Sync?
Enable the Prediction Sync option by clicking the "chasing arrows" icon in the Predictions section of your Project, or by clicking the SETTINGS button.
You can specify the name you'd like for the predicted dataset; by default, the name will be the name of your apply dataset with "_PREDICTED" appended to it. If you choose a name that already exists in your target environment, AutoML will warn you and allow you to choose a new, unique name. AutoML will then immediately write the prediction dataset.
After that, AutoML will keep your predictions up to date by reading the apply dataset on a scheduled basis and writing the prediction dataset to your target environment.
IMPORTANT NOTE: If you disable Prediction Sync, the Keep Up To Date option remains enabled until you disable it. If you don't disable Keep Up To Date, AutoML will continue to generate new predictions on a scheduled basis but will NOT push predictions to your target environment, unless or until you re-enable Prediction Sync.
How can I see the run history and status of my scheduled predictions?
At any time you can access the prediction settings screen and scroll down to see the run history of your predictions. This history will include information on both scheduled predictions as well as those run manually. If you see an error icon, hovering over that icon will provide details about what went wrong with a particular prediction.
Additionally, AutoML displays both Keep Up To Date and Prediction Sync status directly in the Project Overview screen, so that you always know the status of your predictions. Grey icons indicate a disabled status; green icons indicate an enabled status.
If AutoML encounters a problem with either Keep Up To Date or Prediction Sync settings (such as if your database login credentials become out-of-date or if an apply dataset for a synced prediction is no longer available) AutoML now provides a visual indication by turning the icons orange. Hovering over the "warning triangle" icon will give you more detail about the error AutoML encountered.
How do I change the name of my synced prediction dataset?
Open prediction settings, then click on the current name of your synced prediction dataset. As soon as you enter new text or change the existing text, AutoML will prompt you to confirm that you want to change the name. Clicking OK will enable the new name, which AutoML will begin using the next time your predictions are synced, either through your prediction schedule or by manually running the predictions. Clicking Cancel will discard your changes.
When and how often will AutoML update my predictions?
By default, AutoML will run your predictions on a daily basis, between 2am and 6am Mountain Time. If you would like to choose a specific time for all of your daily scheduled predictions, contact AutoML Support using the link at the top of your screen and we will set a daily time for your predictions that you choose. Note that changing the default time will affect all predictions for your user ID. If you want to specify different times for different predictions, check out the "custom schedule" section, directly below.
Can AutoML run my scheduled predictions more than once a day or on a custom schedule?
Big Squid customers with an Enterprise plan have the ability to request unlimited custom schedules for prediction syncs. Need your prediction to run every hour between 8am and 10pm on weekdays plus twice a day on weekends? AutoML can do that. Want to run a prediction every other day at 10am and 2pm, plus every hour on the first day of the month? AutoML can do that.
Enterprise plan customers can contact their friendly neighborhood Big Squid Customer Success Representative or AutoML Support to request the creation of custom prediction schedules. If you are not an Enterprise plan customer, please Qlik Sales for more information.
Important considerations about Prediction Sync:
- Ensure that the synced dataset name you plan to use does not already exist in your target environment. If the name already exists, AutoML will be unable to write the prediction dataset and will generate an error. This is by design, to prevent you from inadvertently overwriting existing data in your target environment.
- Related to - but distinct from - consideration #1 above, once you have created a unique prediction dataset name, AutoML will overwrite that prediction dataset during scheduled predictions as well as if you click the "Run predictions now" button in the prediction settings. In other words, AutoML does not append predictions but, instead, overwrites the previous prediction dataset.
- Check out this article for information on using dynamic prediction dataset naming, which allows you to include a variety of time-based variables that AutoML will replace with actual values when your prediction dataset is synced.
How is Prediction Sync different from Automated Prediction Download?
Prediction Sync writes AutoML predictions on a scheduled basis for any apply dataset EXCEPT those powered by CSV file upload. AutoML writes the data to your target environment on a scheduled basis in what's commonly referred to as a data "push".
Conversely, Automated Prediction Download is specifically designed to allow you to retrieve ("pull") prediction datasets that AutoML has already generated from any apply dataset, including manually-uploaded .CSV data. It is a "one-way street" from AutoML to your applications to access existing predictions at any time you choose. It's important to note that, while you can use Automated Prediction Download to pull predictions from AutoML at any time, the "freshness" of those predictions will be based on the last time the predictions were updated in AutoML, either manually by a user or automatically if the Keep Up To Date option is enabled for your predictions. In other words, Automated Prediction Download does not force AutoML to refresh the predictions; it simply provides access to the latest existing predictions.
There are many reasons why you may choose one or the other - as well as when you may want to use both - for a given prediction dataset. See the "Why (and how) would I use Automated Prediction Download?" section of the Automated Prediction Download documentation for more information.
Are there any other Prediction Sync resources I should know about?
Check out the legacy video for a brief overview of how Prediction Sync works and why you would use it, as well as how it relates to Keep Up To Date.