- Newest
- Most votes
- Most comments
Hi, Amazon Forecast automatically scales up or down in response to forecasting and data volume requirements. When selecting the AutoML option, however, there is no method to manually change the instance size or type. The AutoML option selects the best possible algorithm and hyperparameters for your data automatically, however it may take longer to train than a predetermined algorithm.
If you want more control over the training duration and performance, you may choose one of the preconfigured algorithms provided by Amazon Forecast, such as ARIMA, DeepAR+, ETS, NPTS, or Prophet3. You may also select the prediction horizon, forecast frequency, and quantiles for your predictor. For example, you can select up to 5 quantiles between 0.01 and 0.99 (in increments of 0.01). These settings may have an impact on the training duration and accuracy of your predictor.
Alternatively, you could use the AWS CloudFormation and AWS Step Functions to create repeatable and sustainable forecasting schedules. This allows you to automate the process of developing predictors and predictions, as well as track their development and performance. AWS Lambda functions can also be used to activate actions based on certain events or situations.
I hope this helps you with your question. 😊
Relevant content
- Accepted Answerasked 3 years ago
- asked 4 years ago
- Accepted Answerasked 4 years ago
- AWS OFFICIALUpdated 7 months ago
- AWS OFFICIALUpdated 2 years ago
- AWS OFFICIALUpdated 8 months ago
- AWS OFFICIALUpdated a year ago