Posted On: Jan 8, 2018
You can now use the DeepAR forecasting algorithm for model training in Amazon SageMaker.
DeepAR is an algorithm that generates accurate forecasts by learning patterns from time-series over multiple large sets of training data with related time-series. The DeepAR algorithm learns similarities across the related items in the dataset to provide more accurate forecasts. This method improves upon common forecasting methods like Autoregressive Integrated Moving Average (ARIMA) models or exponential smoothing which treat each time-series independently. By using the shared information across related time-series, DeepAR can be applied to a number of time-series challenges, such as predicting future product sales to improve supply chain management, forecasting traffic to servers or web pages, or estimating future electricity consumption at the individual household level.
The DeepAR algorithm is available today in the US East (N. Virginia & Ohio), EU (Ireland) and U.S. West (Oregon) AWS regions. To learn more, visit the Amazon SageMaker documentation for the DeepAR algorithm.