The DeepAR algorithm offered by Sagemaker is a generalized deep learning model that learns about demand across several related time series. Unlike traditional forecasting methods, in which an individual time series is modeled, DeepAR models thousands or millions of related time series.
Examples include forecasting load for servers in a data center, or forecasting demand for all products that a retailer offers, and energy consumption of individual households. The unique thing about this approach is that a substantial amount of data on past behavior of similar or related time series can be leveraged for forecasting an individual time series. This approach addresses over-fitting issues and time—and labor-intensive manual feature engineering and model selection steps required by traditional techniques.
DeepAR is a forecasting method based on autoregressive...
