Microsoft Office Tutorials and References
In Depth Information
Exhibit 3.28 Correlation matrix for Product A–E
3.5.4 Other Forecasting Models
In a more in-depth investigation of the data, we would include a search for other
appropriate models to describe the data behavior. These models could then be used
to predict future quarterly periods. Forecasting models and techniques abound and
require very careful and studied analysis, but a good candidate model for this data
is one that is known as Winters’ 3-factor exponential smoothing . The conceptual
fit appears to be excellent. Winters’ model assumes 3 components in the structure
of a forecast model—a base or level, a linear trend, and some form of cyclicality.
All these elements appear to be present in most of the data series for product sales
and are also part of our previous analytical assumptions. The Winters’ model also
incorporates the differences between the actual and predicted values (errors) into its
future calculations: that is, it incorporates a self-corrective capability to account for
errors made in forecasting. This self-corrective property permits the model to adjust
to changes that may be occurring in underlying behavior. A much simpler version of
Winters’ model is found in Data Analysis as Exponential Smoothing , which only
assumes a base or level component of sales.
Search JabSto ::

Custom Search