Join Conrad Carlberg for an in-depth discussion in this video Next steps, part of Business Analytics: Forecasting with Seasonal Baseline Smoothing.
- [Conrad] There are plenty of features used…by different smoothing models…that this course hasn't had enough time…to explore even briefly, let alone in depth.…If you intend to look further…into smoothing as an approach to forecasting,…I urge you to examine a model not only…from the point of view of the summary statistics provided…by functions in R,…but also the period to period details…that show most clearly in an Excel worksheet.…There are so many choices involved…in specifying a smoothing model…that it can be very difficult to distinguish…a change in the nature of the baseline…from an apparently minor change…in the way a model is specified.…
Those choices can involve differences…between trended and seasonal models,…add a difference in small duplicative models,…whether a damping parameter has been added to the equation…and which error measure has been chosen…to evaluate the smoothing constants.…With that in mind, I encourage you…to explore some of my other courses in this library.…
- Identify what distinguishes seasonality from a trend or a cycle.
- Explore how to use absolute and relative references in defined names, and recall that absolute reference always remain static while relative references change depending on precedent.
- Identify seasonality in a baseline by examining autocorrelation functions in a correlogram.
- Explore how to initialize seasonal effects in a baseline.
- Forecast the current level of the baseline and the current seasonal effect from prior observations, forecasts, and smoothing constants.
- Quantify a measure of the aggregate error in a forecast, and minimize it using Solver.
- Establish a baseline in a data object and forecast from that baseline in R.
- Compare Excel and R results.