Imputed variable generator

• US 7,062,447 B1
• Filed: 12/20/2000
• Issued: 06/13/2006
• Est. Priority Date: 12/20/2000
• Status: Expired due to Term

• Current Assignee: Acoustic, L.P.
• Original Assignee: DemandTec Inc. (International Business Machines Corporation)
• Inventors: Venkatraman, Krishna, Delurgio, Phil, Neal, Michael, Valentine, Suzanne, Lee, Hau
• Primary Examiner(s): Backer, Firmin
• Assistant Examiner(s): Greene, Daniel L.
• Application Number:
  US09/742,472
• Time in Patent Office: 2,001 Days
• Field of Search: 705/400, 705/40, 705/1, 705/20
• US Class Current: 705/1.1
• CPC Class Codes: G06Q 10/00 Administration; Management
• CPC Class Codes: G06Q 20/102 Bill distribution or payments
• CPC Class Codes: G06Q 20/201 Price look-up processing, e.g. updating
• CPC Class Codes: G06Q 30/0206 Price or cost determination based on market factors

Abstract

Disclosed are methods, media and systems for generating imputing econometric variables from raw econometric data. A method for generating cleansed initial dataset from corrected raw econometric data comprises inputting raw econometric data, formatting and classifying the raw econometric, conducting an initial error detection and correction, defining a store data set hierarchy, conducting a second error detection and correction, defining product attributes and demand groups, updating attribute information, defining equivalizing factors, conducting a third error detection and correction, subsetting data to facilitate increased processing speeds, conducting a fourth error detection and correction and outputting the cleansed initial dataset. Another embodiment, capable of generating imputing econometric variables from corrected raw econometric data comprises receiving the raw econometric data, detecting errors in the raw econometric data, correcting the detected errors in the raw econometric data to generate a cleansed initial dataset, and generating imputed econometric variables using the cleansed initial dataset.

Claims

1. A method for imputing econometric variables from corrected raw econometric data, the method being implemented as a plurality of program instructions in a computer system, the method comprising:

detecting, using the computer system, product parameter inconsistencies in the raw econometric data;

correcting, using the computer system, the detected inconsistencies in the raw econometric data to generate a cleansed initial dataset; and

generating, using the computer system, imputed econometric variables using the cleansed initial dataset.

2. The method of claim 1, wherein the imputed econometric variables include imputed variables selected from the group consisting of an imputed base price variable, an imputed relative price variable, an imputed base volume variable, an imputed variable reflecting the effects of stockpiling, an imputed variable reflecting seasonal effects, an imputed variable reflecting day-of-the-week effects, an imputed variable reflecting promotional effects, and an imputed cross-elasticity variable.

3. The method of claim 2, wherein the imputed econometric variables include imputed variables selected from the group consisting of an imputed base price variable, an imputed relative price variable, an imputed base volume variable, an imputed variable reflecting the effects of stockpiling, an imputed variable reflecting day-of-the-week effects, an imputed variable reflecting promotional effects, and an imputed cross-elasticity variable.

4. The method of claim 1, wherein imputing econometric variables include imputing a base price variable.

5. The method of claim 4, wherein imputing base price variable comprises:

defining a time window for analyzing records of the cleansed initial dataset;

determining an initial base based on data within the time window; and

outputting the initial base price as the base price variable.

6. The method of claim 4, wherein imputing base price variable comprises:

defining a time window for analyzing records of the cleansed initial dataset;

determining an initial base price based on data within the time window;

determining promotional effects on the initial base price based on data within the time window;

defining a step function for initial base price based on data within the time window;

calculating price discount variables;

outputting initial base price step function and discount variables;

analyzing and correcting the outputted initial base price variable based on base price distribution data and generating a refined base price variable; and

outputting the refined base price variable.

7. The method of claim 1, wherein imputing econometric variables includes imputing a relative price variable.

8. The method of claim 7, wherein imputing relative price variable comprises:

determining equivalent price for price values in the cleansed initial dataset;

determining equivalent units for units values in the cleansed initial dataset;

calculating values for equivalent base price and equivalent base units;

determining a weighted equivalent average price;

determining a weighted equivalent average base price;

calculating moving averages for relative equivalent price and relative equivalent base price; and

outputting relative equivalent price and relative equivalent base price as relative price variables.

9. The method of claim 1, wherein imputing econometric variables includes imputing a base volume variable.

10. The method of claim 9, wherein imputing base volume variable comprises:

analyzing the cleansed initial dataset to determine non-promoted dates;

calculating average sales volume during the non-promoted dates;

determining an initial units value based on the cleansed initial dataset and the calculated average sales volume during the non-promoted dates;

determining a moving average value for initial units; and

outputting the moving average value for initial units as the imputed base volume variable.

11. The method of claim 1, wherein imputing econometric variables includes imputing a variable reflecting the effects of stockpiling.

12. The method of claim 1, wherein imputing econometric variables includes imputing a variable reflecting day-of-the-week effects.

13. The method of claim 12, wherein imputing variable reflecting day-of-the-week effects comprises:

assigning each record of the cleansed initial dataset to a specific day of the week;

specifying an input dimension;

summing each assigning record over the specified input dimension;

determining, for each specific day of the week, an average amount of units sold for each day; and

determining for each day of the week, using actual daily units sold divided by average amount of units sold for each day, a relative daily volume of units sold variable.

14. The method of claim 1, wherein imputing econometric variables includes imputing variable reflecting promotional effects.

15. The method of claim 14, wherein imputing variable reflecting promotional effects comprises:

using cleansed initial dataset and calculated values for base units, determining crude promotional variable;

conducting simple regression analysis to obtain initial volume model;

comparing sales volume results generated by the initial volume model with actual sales volume; and

adjusting initial volume model to reflect inconsistencies between the initial volume model and the actual sales volume to generate a corrected volume model, which comprises an imputed, promotional affects variable.

16. The method of claim 1, wherein imputing econometric variables includes imputing a cross-elasticity variable.

17. The method of claim 16, wherein imputing cross-elasticity variable comprises:

inputting demand group information from the cleansed initial dataset;

calculating equivalent sales volume for each demand group per time period;

calculating average equivalent sales volume for each demand group per time period;

calculating relative equivalent sales volume for each demand group per time period; and

generating cross-elasticity variables for each demand group.

18. A computer-readable medium having programming instructions arranged to impute econometric variables from raw econometric data, the computer readable medium including:

programming instructions for receiving the raw econometric data;

programming instructions for detecting product parameter inconsistencies in the raw econometric data;

programming instructions for correcting the detected inconsistencies in the raw econometric data to generate a cleansed initial dataset; and

programming instructions for generating imputed econometric variables using the cleansed initial dataset.