Amazon Forecast provides a fully managed service that applies statistical and machine learning algorithms to generate time-series forecasts from historical data. It automates the forecasting process, including model selection and handling of missing values, so users prepare datasets and interpret predictions. The service uses technology developed for Amazon.com forecasting, supports APIs, AWS CLI, SDKs, and console for importing data, training predictors, and creating forecasts. It incorporates built-in datasets and related time series for improved accuracy without machine learning experience required.
Automates model selection to improve accuracy up to 40% over traditional methods
Handles missing values and data preparation automatically
Incorporates weather data at no extra cost for better demand forecasts
Requires data import and preparation as initial steps
Forecasts limited to time horizons based on historical data length
Compute costs apply to training and inference operations
