Academic journal article The Journal of Business Forecasting Methods & Systems

What to Look for in a Forecasting Engine

Academic journal article The Journal of Business Forecasting Methods & Systems

What to Look for in a Forecasting Engine

Article excerpt

This paper outlines a number of important issues that should be considered when acquiring or developing a "forecasting engine." Good management begins with a plan. Proper planning begins with an accurate forecast. Tactical planning calls for solid production management functions with an emphasis on detailed forecasts. Forecasts at this level provide an initiative to reduce inventory costs, cut obsolescence, and eliminate missed opportunities.

The preset models found in the current systems offer suboptimal forecasts. In order to improve upon the current state, the system needs to automatically identify the lead and lag structures and best weights for each of the input series. The need to identify and incorporate omitted variables, pulses, seasonal pulses, level shifts, and local time trends. The variables that are not necessary must also be deleted from the model. The system should report the statistical tests used to create the model while allowing the user to modify the model if they so choose.

An engine that operates automatically and feeds from and into a company's database is clearly useful. Even where a series warrants the analyst's personal attention, the automatically generated forecast will be a good place to start. This paper is presented in a Q & A format. Figure 1 can be used to rate your engine. There are five main issues to consider when evaluating a forecasting engine:

1. Computational considerations

2. Ease of use

3. Model robustness

4. Empowering the user

5. Special items

COMPUTATIONAL CONSIDERATIONS

To streamline the forecast process, it is important for a user to have the maximum flexibility in platform selection, the analytical techniques employed and output control. Allowing the user maximum control combined with an automated approach to model building empowers the user with the capability to identify the most efficient process for generating forecasts.

Q. How does the forecasting engine operate and perform on different platforms? Does the desktop (PC) version give the same answer as the workstation or mainframe version?

A. It should be identical. The engine version must be capable of running on a large number of platforms.

Q. Does the forecast engine fit neatly into an existing decision support system (MIS)? How does it import data and report results? Can specific targeted information be processed subsequently?

A. The engine must be able to interface into other systems seamlessly. The engine should read and write flat to ASCII files. Alternatively, data should be capable of being loaded via ODBC facilitating an

Q. Does the forecast engine deliver forecasts and statistical results such as RI, model equation, an audit trail of the expert systems process, and table of forecasts? Does the engine provide such output in user defined files?

A. The forecasts and steps to build the model must be displayed so that the user can understand how the expert system arrived at its answer. The engine must allow the user to specify file names where forecasts are saved. Files should be saved so that they can be imported into other applications. Statistical information must be directed to specific files. Segmenting statistics into different files allows the user (for example) to get a histogram of the R2 for all of the series at once.

Q. Is there an approach which permits the user to speed up the process?

A. To gain computational speed the engine should allow the user to disable/enable various tests or increase/decrease the maximum number of iterations allowed. The user must also be able to control various print options or the level of output verbosity. This control will help to speed up the process if the additional print out is not necessary.

EASE OF USE

The forecasting engine should not only have the automatic process of model identification and forecasting, but also be user friendly. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.