Peng Liu, PhD, JMP Principal Research Statistician Developer, SAS
Jian Cao, PhD, JMP Principal Systems Engineer, SAS

A new Time Series Forecast platform has been developed to enable users to build forecasting models for multiple time series easily and fast. In comparison to the conventional univariate time series forecasting models, this new platform streamlines the model building process with minimal user input and makes forecasts for hundreds or thousands of time series in a single run. This is achieved by employing flexible modern smoothing methods and a well-accepted model selection procedure. Research has shown this approach's forecasting accuracy is compelling.

In this paper we will discuss the statistical details behind the new forecast platform and demonstrate real-life examples to show how the implementation achieves computational efficiency.