Amazon Forecast
Amazon Forecast is a machine learning service from Amazon Web Services that produces forecasts from time-series data — future demand, inventory levels, staffing needs, financial or capacity planning — without requiring the user to build a model.
AI Practitioner focus
- Forecast is useful historical service context, but it is closed to new customers and is not on the current AIF-C01 in-scope list. Use SageMaker Canvas or another current forecasting approach for new workloads.
- Do not let the older supplied study sheet turn Forecast into a current-scope distractor.
Key points
- Takes historical time-series data plus any related variables (price, promotions, weather, holidays) and selects a suitable algorithm automatically.
- Aimed at business planning questions: how much stock to hold, how many staff to schedule, what resource demand to expect.
- Requires no machine learning expertise — a pre-trained, managed service rather than a model built in Amazon Sage Maker.
- Closed to new customers as of July 29, 2024. Existing customers can keep using it and AWS continues security, availability, and performance updates, but no new features are planned. AWS recommends new forecasting workloads use Amazon SageMaker Canvas, which AWS states builds models up to 50% faster and produces predictions about 45% quicker on average.
- The boundary that matters: Forecast predicts future values of a series. Amazon Personalize recommends items to individual users. Amazon Timestream is the database for storing and querying time-series data, not for projecting it forward.