Demand forecasting method, demand forecasting system, and recording medium

• US 6,032,125 A
• Filed: 10/22/1997
• Issued: 02/29/2000
• Est. Priority Date: 11/07/1996
• Status: Expired due to Term

• Current Assignee: Fujitsu Limited
• Original Assignee: Fujitsu Limited
• Inventors: Ando, Hideyuki
• Primary Examiner(s): Trammell, James P.
• Assistant Examiner(s): Nguyen, Cuong H.
• Application Number:
  US08/955,663
• Time in Patent Office: 860 Days
• Field of Search: 705/10, 705/14, 705/28, 705/7, 707/2, 702/3, 706/45, 706/14, 706/15, 706/931, 706/20-21, 379/113, 379/133, 379/309, 379/352
• US Class Current: 705/7.31
• CPC Class Codes: G06Q 10/087 Inventory or stock management, e.g. order filling, procurement or balancing against orders
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: G06Q 30/0202 Market predictions or forecasting for commercial activities

Abstract

A method and a system for forecasting the demand agreeing with the fluctuation trend of sales results at high and stable precision, without requiring user'"'"'s maintenance, by using a model optimum for grasping the fluctuation trend of sales results, even if the products are diverse, by storing a plurality of models of neural network, for example, a model for forecasting the demand from data of the past several months, a model for forecasting the demand from data of the same period of the previous year, and a model for forecasting the demand from both the latest data and data of the same period of the previous year, and also by feeding sales results into a model of neural network to make it learn by the short period such as by the week, and a recording medium in which is recorded such program.

Claims

1. A method for forecasting demand for a product based on sales results of the product, comprising:

setting plural models as a neural network;

identifying sales results of a first period;

inputting the identified sales results of the first period to each of the models to make the neural network of each model learn from inputs and produce data as close as possible to sales results of a second period following the first period;

storing a forecast demand value of a predetermined time outputted by each of the neural networks;

selecting a model from the learned neural networks which has a forecast demand value closest to the sales results of the predetermined time; and

inputting latest sales results identified by the learned neural network corresponding to the selected model to forecast demand.

2. A demand forecasting method of claim 1, further comprising an outputting device outputting a calculated error between the sales results and demand forecasting result.

3. A method of claim 1, wherein said model is a model incorporating position data indicating period position on a calendar as a processing element, and the position data is fed in the neural network together with the sales results.

4. A method of claim 1, wherein said model is a model incorporating the position data indicating the position of a calendar period as a processing element, and the position data is fed in the neural network together with the sales results.

5. A demand forecasting system using the method of claim 4.

6. A demand forecasting system of claim 5, further comprising an output device outputting a calculated error between the sales results and demand forecasting result.

7. A method of claim 1, wherein the sales results used in learning of the neural network of 13 months dating back from a learning point is acquired.

8. A demand forecasting system using the method of claim 7.

9. A demand forecasting system of claim 8, further comprising an output device outputting the error between the sales results and demand forecasting result.

10. A method of claim 1, wherein demand forecasting in a first period unit forecast by the neural network is reflected in the demand forecasting in a second period unit composed of a set of first period units.

11. A demand forecasting system using the method of claim 10.

12. A demand forecasting system of claim 11, further comprising an output device for outputting the error between the sales results and demand forecasting result.

19. A recording medium of claim 5, further comprising reflecting the demand forecasting in a first period unit forecast by the neural network in the demand forecasting in a second period unit composed of a set of the first period units.

13. A computer readable storage media storing a process of forecasting the demand for a product on the basis of the sales results of the product comprising:

setting plural models as a neural network;

identifying sales results of a first period;

inputting the identified sales results of the first period to each of the models to make the neural network of each model learn from inputs and produce data as close as possible to the sales results of a second period following the first period;

storing a forecast demand value of a predetermined time outputted by each of the neural networks;

selecting a model from the learned neural networks which has a demand value closest to the sales results of the predetermined time; and

inputting a latest sales results identified by the learned neural network corresponding to the selected model to forecast a demand.

14. A recording medium of claim 13, wherein said model is a model incorporating a position data indicating position of a calendar period as a processing element, and further including program code means for causing said computer to feed the position data in the neural network together with the sales results.

15. A recording medium of claim 13, wherein the identifying sales results includes acquiring results of 13 months dating back from the learning point.

16. A recording medium of claim 13, further including causing said computer to reflect the demand forecasting in a first period unit forecast by the neural network in the demand forecasting in a second period unit composed of a set of the first period units.

17. A recording medium of claim 14, wherein identifying sales results includes acquiring results of 13 months dating back from the learning point.

18. A recording medium of claim 14, further comprising reflecting the demand forecasting in a first period unit forecast by the neural network in the demand forecasting in a second period unit composed of a set of the first period units.

20. A demand forecasting method comprising:

creating a plurality of neural network models to forecast demand based on different time periods;

identifying sales results of a first period and entering the results into each of the models to allow each model to learn and forecast demand for a second period;

comparing the forecast demand from each of the models for the second period with actual sales results to compute an error of each model; and

selecting the model with the smallest error.

21. A computer readable storage medium storing software to implement a demand forecasting method performing;

creating a plurality of neural network models to forecast demand based on different time periods;

identifying sales results of a first period and entering the results into each of the models to allow each Model to learn and forecast demand for a second period;

comparing the forecast demand from each of the models for the second period with actual sales results to compute an error of each model; and

selecting the model with the smallest error.

22. A demand forecasting system comprising:

neural network models forecasting demand based on different time periods;

an inputting device inputting sales results of a first period into each of the models;

a comparing device comparing a forecast demand from each of the models for a second period with actual sales results to compute an error of each model; and

a selecting device selecting the model with the smallest error.