Method and system for combination of independent demand data streams

Patent metadata

Field	Value
Publication number	US-9747560-B2
Application number	US-18165305-A
Country	US
Kind code	B2
Filing date	Jul 13, 2005
Priority date	Jul 13, 2005
Publication date	Aug 29, 2017
Grant date	Aug 29, 2017

Abstract

A system and method of combining independent demand forecast streams. Demand data is extracted from each of the demand forecast streams. The extracted data is combined based on one or more criteria to yield a single resultant demand data set. The resultant demand data set is released for supply network planning rather than any original forecast data stream.

First claim

What is claimed is:

1. A method comprising:
extracting, in a computer, first demand data from a first forecast demand data stream, wherein the first demand data stream is of a first type;
extracting, in the computer, second demand data from a second forecast demand data stream, wherein the second demand data stream is of a second type different from the first type and wherein both demand data streams apply to a same source of demand;
storing the extracted first demand data and the extracted second demand data in a database of a data warehouse;
at the computer, computing and storing parameters associated with an error measurement for the first demand data and for the second demand data at a demand combination info cube;
determining correspondingly, by the computer, frequencies of data extraction for demand data combination from first demand data and second demand data;
combining, in the computer, the first demand data and the second demand data based on at least one criteria associated with the stored parameters, to generate the demand combination info cube, wherein generating the demand combination info cube includes combining the first and second demand data based on:
evaluating, by a rule engine, characteristics combinations of the first demand data and the second demand data by traversing conditions defined at nodes of a decision tree, wherein the conditions are associated with comparisons of forecast key figures defined in the first and second demand data for a time interval;
calculating a resultant demand data set based on performing a sequence of demand combination activities, wherein the sequence of demand combination activities is determined during traversing the decision tree, and wherein the resultant demand data set is to be stored at the demand combination info cube;
determining modified frequencies of data extraction based on the sequence of demand combination activities; and
performing a modification, by the computer, of the frequencies of data extraction from the first demand data and the second demand data, based on the calculated resultant demand data set; and
releasing the demand combination info cube for supply network planning.

2. The method of claim 1 wherein combining comprises:
calculating the parameters to include error measurement statistics for the first and the second demand data; and
selecting demand data from the demand data having a least error as part of the resultant demand data set.

3. The method of claim 2 wherein calculating comprises:
determining one of absolute percentage error (APE) and weighted absolute percentage error (WAPE).

4. The method of claim 1 wherein combining comprises:
applying a first weighting factor to the first demand data to obtain a first weighted demand data;
applying a second weighting factor to the second demand data to obtain a second weighted demand data; and
summing the first weighted demand data with the second weighted demand data.

5. The method of claim 1 further comprising:
normalizing the first and second demand data to correspond to the time interval.

6. The method of claim 1 further comprising:
comparing one of the first demand data and the second demand to previous demand data from a corresponding one of the first forecast demand data stream and the second forecast demand data stream; and
updating the combining if the comparing shows a difference exceeding a threshold.

7. The method of claim 1 wherein traversing comprises:
evaluating one or more binary conditions from the conditions defined in the decision tree to identify an activity part of the sequence of demand combination activities; and
performing the activity to generate a member of the resultant demand data set.

8. A system comprising:
a processor;
a first extractor to extract first demand data from a first forecast demand data stream, the first demand data stream being of a first type;
a second extractor to extract second demand data from a second forecast demand data stream, the second demand data stream being of a second type different from the first type;
a database of a data warehouse to retain historical demand data and the extracted first demand data and the extracted second demand data;
at the data warehouse, compute and store parameters associated with error measurement for the first demand data and the second demand data at a demand combination info cube;
determine correspondingly, by the computer, frequencies of data extraction for demand data combination from first demand data and second demand data; and
a combination module to combine the first and second demand data into a single resultant data set and to generate the demand combination info cube, wherein determining the single resultant data set is based, at least in part, on evaluations associated with the historical demand data,
and wherein both demand data streams apply to a same source of demand,
and wherein to determining the demand combination info cube includes combining the first and the second demand data based on instructions to:
evaluate, by a rule engine, characteristics combinations of the first and the second demand data by traversing conditions defined at nodes of a decision tree, wherein the conditions are associated with comparisons of forecast key figures defined in the first and second demand data for a time interval;
calculate the single resultant demand data set based on performing a sequence of demand combination activities, wherein the sequence of demand combination activities is determined during traversing the decision tree; and
determine modified frequencies of data extraction based on the sequence of demand combination activities; and
perform a modification, by the computer, of the frequencies of data extraction from the first demand data and the second demand data, based on the calculated resultant demand data set.

9. The system of claim 8 further comprising:
a demand planning module to provide the first forecast demand data stream; and
a responsive replenishment module to provide the second forecast demand data stream.

10. The system of claim 8 further comprising an analysis module, wherein the analysis module comprises:
the rule engine to evaluate the first and the second demand data against the conditions defined in the nodes of the decision tree.

11. The system of claim 10 further comprising:
a graphical user interface to permit modification of the set of rules.

12. The system of claim 10 wherein the analysis module comprises:
an error calculation component.

13. A non-transitory machine-accessible medium containing instructions that when executed cause a machine to:
extract first demand data from a first forecast demand data stream, wherein the first demand data stream is of a first type;
extract second demand data from a second forecast demand data stream, wherein the second demand data stream is of a second type different from the first type and wherein both demand data streams apply to a same source of demand;
store the extracted first demand data and the second demand data in a database of a data warehouse;
at the data warehouse, compute and store parameters associated with error measurement for the first demand data and the second demand data at a demand combination info cube;
determine correspondingly, by the computer, frequencies of data extraction for demand data combination from first demand data and second demand data;
combine the first and second demand data based on at least one criteria associated with the stored parameters to generate the demand combination info cube, wherein the instructions to determine the demand combination info cube inc

Assignees

• John Thomas
• Traeger Konrad
• Eftekhari Newsha
• Sap Se

Inventors

• John Thomas
• Traeger Konrad
• Eftekhari Newsha

Classifications

• G06Q10/06 Primary

  Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling · CPC title

• G06Q30/0202

  Market predictions or forecasting for commercial activities · CPC title