Method for determining optimal time series forecasting parameters

• US 6,611,726 B1
• Filed: 09/15/2000
• Issued: 08/26/2003
• Est. Priority Date: 09/17/1999
• Status: Expired due to Fees

• Current Assignee: Carl E. Crosswhite
• Original Assignee: Carl E. Crosswhite
• Inventors: Crosswhite, Carl E.
• Primary Examiner(s): Gordon, Paul P.
• Application Number:
  US09/663,401
• Time in Patent Office: 1,075 Days
• Field of Search: 700/52, 700/97, 700/99, 700/100, 700/106, 700/107, 702/180, 702/181, 705/28, 705/29, 705/8
• US Class Current: 700/99
• CPC Class Codes: G06F 17/18 for evaluating statistical data , e.g. average values, frequency distributions, probability functions, regression analysis forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04
• CPC Class Codes: G06Q 10/06 Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
• CPC Class Codes: G06Q 10/087 Inventory or stock management, e.g. order filling, procurement or balancing against orders

Abstract

A method for forecasting a value of a dependent variable, such as product demand, in a future time period later than the next, upcoming future time period. The method includes selecting a dependent variable for which a value is to be forecast, gathering historical data on values of the dependent variable and explanatory variables in prior time periods, and determining a forecasting equation based on the gathered historical data. The method includes selecting a future time period that is a number of time periods beyond the next, upcoming time period. The forecasting method continues with calculating a forecasted value of the dependent variable for the selected future time period, then determining an error value by comparing the forecasted value with the historical data and based on the error value, modifying the forecasting equation to reduce the error value. The forecasting equation may be a time series forecasting equation and the determining of the forecasting equation includes initial setting values for included time series forecasting parameters. The modifying of forecast equation then includes adjusting these forecasting parameters to lower or otherwise optimize the error value. Particularly, the method includes selecting an error metric for optimization for the forecasting equation and the adjusting of the parameters is performed as a function of the selected error metric to move it toward an optimal value.

Claims

1. A method of forecasting a value of a dependent variable, such as product demand, in a future time period later than the next, upcoming future time period, comprising:

selecting a dependent variable for which a value is to be forecast;

gathering historical data on values of the dependent variable and explanatory variables in prior time periods;

determining a forecasting equation based on the gathered historical data;

selecting a future time period that is a number of time periods beyond the next, upcoming time period;

with the forecasting equation, calculating a forecasted value of the dependent variable for the selected future time period;

determining an error value by comparing the forecasted value with the historical data; and

based on the error value, modifying the forecasting equation to improve the error value.

2. The method of claim 1, wherein the forecast equation determining includes initially setting values for forecast parameters and wherein the modifying of the forecasting equation comprises adjusting the values of the forecast parameters.

3. The method of claim 2, wherein the forecasting equation is a time series forecasting equation and the forecast parameters are time series forecasting parameters.

4. The method of claim 2, further including selecting an error metric for the forecasting equation for optimization and wherein the adjusting of the values of the forecast parameters is performed to move the error metric toward a lower value.

5. The method of claim 4, wherein the error value determining comprises accumulating error values as a function of the error metric.

6. The method of claim 5, wherein the error metric is selected from the group of error metrics consisting of mean absolute deviation (MAD), mean absolute error (MAE), mean absolute percentage error (MAPE), and root mean squared error (RMSE).

7. The method of claim 1, further including prior to the calculating, cleansing the gathered historical data comprising identifying and replacing missing data in the gathered historical data and identifying and modifying outliers in the gathered historical data.

8. The method of claim 1, further including determining whether the error value is optimized and if optimized, skipping the modifying of the forecast equation and ending the method.

9. The method of claim 1, wherein the selecting includes selecting at least a second future time period and assigning an importance value to each of the selected future time periods.

10. The method of claim 9, wherein the calculating and determining of the error value are performed for each of the future time periods, the determining including multiplying the importance values to the error values.

11. A computer program for forecasting embodied in a tangible media, comprising:

first computer code devices configured to cause a computer to request and accept as input a dependent variable to be used in the forecasting;

second computer code devices configured to cause a computer to gather and store in memory historical data for the dependent variable and for explanatory variables;

third computer code devices configured to cause a computer to determine a forecasting equation based on the stored historical data;

fourth computer code devices configured to cause a computer to request and accept as input a future time period for performing the forecasting;

fifth computer code devices configured to cause a computer to calculate with the forecasting equation a forecasted value for the dependent variable in the inputted future time period;

sixth computer code devices configured to cause a computer to determine an error value based on a comparison of the forecasted value and the historical data; and

seventh computer code devices configured to compare the error value with an optimal value and based on the comparison modify the forecasting equation to move the error value to be closer in value to the optimal value.

12. The computer program of claim 11, wherein the forecast equation is a time series forecasting equation and the third computer code devices are configured to cause a computer to determine initial values for forecast parameters in the forecast equation automatically or by receiving input.

13. The computer program of claim 12, wherein seventh computer code devices are configured to cause the computer to modify the forecasting equation by modifying the forecasting parameters.

14. A computer-based method for forecasting a value of a dependent variable in a future time period as a function of values of explanatory independent variables, the future time period being a selectable number of time periods beyond a current time period, comprising:

collecting historical data on the values of the dependent variable and the independent variables;

determining a form of a time series forecasting equation using the historical data;

receiving as input the future time period;

initializing forecast parameters of the time series forecasting equation;

selecting an error metric for the time series forecasting equation;

determining a value for the error metric for the future time period including calculating an error term and accumulating the error term as a function of the error metric;

comparing the value of the error metric to an optimal value; and

adjusting the forecast parameters based on the comparing to move the value of the error metric toward the optimal value.

15. The method of claim 14, further including receiving from a user as input a dependent variable prior to the collecting.

16. The method of claim 14, further including repeating the calculating, the error metric value determining, the comparing, and the adjusting until the value of the error metric is substantially equal to the optimal value.