May 15, 2017: We are excited to announce the release of the latest version of NumXL 1.65 (HAMMOCK).
In this release, we have revamped the NumXL exponential smoothing functions and created a new optimizer for finding optimal smoothing parameters, combining the best stochastic and computational search algorithms. Furthermore, the smoothing function returns a larger selection of output types: out-of-sample, optimal values of smoothing parameters, level/trend/seasonal, and forecast in-sample components series. Finally, we introduced a new function to support all types of trend and/or seasonal components to capture over 10 additional exponential smoothing models beyond our original set, including Brown’s simple, Brown’s linear, Holt’s double, and Holt-Winters multiplicative exponential smoothing functions.
In this version, we have added a whole new category for forecasting performance measures, a key metric not only for comparing different forecasting models within one (or multiple) time series but also for quantifying and tracking forecast process capabilities over time. We have chosen over 15 measures used widely in academic spheres and in practice, such as mean squared error (MSE), median of absolute percentage error (MdAPE), mean relative absolute error (MRAE), mean absolute scaled error (MASE), percentage better (PB), and others.
We have also expanded our online support documentation, and now include more Spanish technical notes and tutorials.
The idea for these updates came from you. As always, we aim to make the analysis process better by providing you with the tools you need to answer more of your data questions.
For a complete list of the changes, please refer to the NumXL 1.65 release notes.