SAS® High-Performance Forecasting

Produce millions of high-quality forecasts quickly and automatically to improve planning and decision making

Features

Data preparation

• Aggregates transactional data into time series format.
• Takes care of missing values.
• Accumulates data into any frequency time series.

Sophisticated modeling techniques

• Automatic outlier detection.
• Trend analysis.
• Optimized model parameters.
• By-group processing for flexible forecasting hierarchies.
• Seasonality and intermittent series tests.
• Handles zero-padded data (leading or trailing zeros recorded in the data).
• Extensible model repository.
• Predefined as well as customizable holiday and calendar event definitions.
• Automatic indicator variable creation.
• Reconciliation of hierarchical forecasts.

Automatic model selection

• Seasonal and non-seasonal exponential smoothing models.
• ARIMAX models.
• Intermittent demand models.
• Unobserved components models.
• User-defined models.
• Automatic variable transformations (log, square root, logistic, Box-Cox).
• Model selection based on performance in a holdout sample region (optional).
• Several goodness-of-fit statistics and overall performance metrics.
• Automatic regressor and event selection and model specification. Determines whether lagged and/or dynamic relationships are present.
• Automatic generation of model selection lists.
• Choice of automation level for all three forecasting steps: model selection, model parameter estimation and forecast generation.

Customized output and reports

• Output produced in several popular formats: HTML, PDF, Microsoft Word, etc.
• SAS ODS graphics (experimental).
• Several output data sets available, including forecasts, goodness-of-fit statistics, model parameters, seasonal and trend statistics, and more.

Superior scalability

• Runtime is linear with the number of forecasts.
• No limit on the amount of historical data or forecast horizon.
• No limit on the number of events or regressors.

Download the complete SAS High-Performance Forecasting Fact Sheet.