Content area
Full Text
Why do businesses need forecasting software packages and systems? To answer this question, we need to know exactly what such packages and systems are and what they do. Before describing that, it is important to understand the difference between forecasting software packages and forecasting systems. Forecasting software is a stand-alone package which is used to generate forecasts using either a forecaster-selected model or an automatic built-in feature (called expert system). The forecasting system, on the other hand, does more than that. It not only has a forecasting engine to prepare forecasts, but also it automates many of the forecasting processes such as acquiring and transmitting forecast data and information.
Most of the forecasting software packages available in the market have an expert system. A package with such a system has an automatic feature that tests a number of forecasting models with the historical data and then prepares forecasts for the next period with a model that, on the average, yields the best results. The criteria used for selecting the best model depend on the type of model selected and the vendor (i.e., the developer of a package). As described in the article on "Benchmarking Forecasting Models," there are three types of models: Time Series, Cause-and-Effect, and Judgmental. Each one has its own tool kit for evaluating a model. In Time-Series models, you may select a model by the lowest Mean Error (ME), Mean Absolute Error (MAE), Mean Absolute Percent Error (MAPE), Mean Squared Error (MSE), or any other criterion. In Regression (which is a part of Cause-and-Effect models), youmay decide to use F value, R2, or any other diagnostic available in its tool kit. The same applies to Judgmental models, though at present, I don't know any vendor that has an expert system based on this type of model. In fact, most of the expert systems available in the market are based just on TimeSeries models; that is to say, their software selects the best model only from a series of Time-Series models. The vendor of a forecasting software package also has its own preference in using the diagnostic tool for selecting the best model. So one forecasting software package may use one tool when selecting the best model, and another may use something else.