Model optimization system using variable scoring

• US 9,147,206 B2
• Filed: 08/05/2010
• Issued: 09/29/2015
• Est. Priority Date: 08/31/2009
• Status: Active Grant

• Current Assignee: Accenture Global Services Limited (Accenture PLC)
• Original Assignee: Accenture Global Services Limited (Accenture PLC)
• Inventors: Umblijs, Andris
• Primary Examiner(s): RIFKIN, BEN M
• Application Number:
  US12/851,391
• Publication Number:
  US 20110055127A1
• Time in Patent Office: 1,881 Days
• Field of Search: 706/12
• US Class Current: 1/1
• CPC Class Codes: G06N 5/025 Extracting rules from data
• CPC Class Codes: G06Q 30/06 Buying, selling or leasing transactions

Abstract

A model optimization system is configured to determine quality of variables for model generation. A data storage stores input variables, quality metrics for the input variables, and weights for the quality metrics. The quality metrics describe sufficiency of data for the input variables and the data is provided for a plurality of regions. A scoring module determines a score for each region based on the input variables and the weighted quality metrics. An optimizer determines whether at least one of the input variables for a region is to be modified based on the scores, and determines whether the total score for the region is operable to be improved using a modified input variable.

Claims

1. A model optimization system to determine quality of variables for model generation, the system comprising:

data storage to store data including input variables, quality metrics for the input variables, and weights for the quality metrics, wherein the quality metrics quantify a sufficiency of the input variables in generating an accurate model, and wherein the data is provided for a plurality of regions;

a processor to execute;

a scoring module to;

determine a quality metric score for each quality metric based on measurements for the quality metric; and

determine an input variable score for each input variable based on the quality metric scores and the quality metric weights;

determine a total score for each region of the plurality of regions based on the input variable scores, wherein to determine the total score for each region, the scoring module is to;

determine categories for the input variables, wherein each category is associated with a type of input variable;

determine category weights for each category;

determine a category score for each category based on the input variable scores for the input variables associated with each category; and

determine the total score for each region based on the category scores and the category weights; and

an optimizer to

determine whether at least one of the input variables for a region of the plurality of regions is to be modified based on the total score for the region, and

in response to a determination that at least one of the input variables for a region of the plurality of regions is to be modified, determine whether the total score for the region is to be improved by applying a modified input variable having a different quality metric.

2. The model optimization system of claim 1, further comprising:

a model builder to generate a model for each region having a total score above a predetermined threshold using the input variables.

3. The model optimization system of claim 1, wherein the categories comprise independent variables controlled by an entity, independent variables outside the entity'"'"'s control, and dependent variables that are dependent on variables in another category.

4. The model optimization system of claim 1, wherein the optimizer is to

identify a total score that falls below a threshold due to at least one of insufficient data from collection sources and inconsistent calculations performed by different sources when determining the data.

5. A method for determining quality of data for modeling, the method comprising:

identifying, by a processor executing computer instructions, input variables to estimate a dependent variable, wherein the input variables are stored in a data storage;

determining quality metrics that quantify a sufficiency of the input variables in generating an accurate model, wherein the data is provided for a plurality of regions;

determining a quality metric score for each quality metric based on measurements for the quality metric;

determining an input variable score for each input variable based on the quality metric scores and quality metric weights for the quality metrics;

determining a total score for each region of the plurality of regions based on the input variable scores, wherein determining the total score for each region comprises;

determining categories for the input variables, wherein each category is associated with a type of input variable;

determining category weights for each category;

determining a category score for each category based on the input variable scores for the input variables associated with each category; and

determining the total score for each region based on the category scores and the category weights;

determining whether at least one of the input variables for a region of the plurality of regions is to be modified based on the total score for the region; and

upon determining that at least one of the input variables for a region of the plurality of regions is to be modified, determining whether the total score for the region is to be improved by applying a modified input variable having a different quality metric.

6. The method of claim 5, wherein determining a quality metric score comprises:

determining a scale of values for each quality metric;

comparing the measurement for each quality metric to a range of values mapped to values within the scale; and

determining the quality metric score for each quality metric based on the comparison.

7. The method of claim 5, wherein the categories comprise independent variables controlled by an entity, independent variables outside the entity'"'"'s control, and dependent variables that are dependent on variables in another category.

8. The method of claim 5, wherein determining the total score for each region comprises:

determining sources for the data for the input variables;

determining source weights for each category; and

determining the total score for each region based on the source weights.

9. The method of claim 5,

wherein the determining of whether at least one of the input variables for a region of the plurality of regions is to be modified comprises determining the at least one of the input variables for the region is to be modified upon a determination that the total score for the region falls below a threshold;

and

wherein the applying a modified input variable having a different quality metric comprises;

identifying an input variable from the input variables operable to be improved based on the quality metrics;

modifying the identified input variable;

re-scoring the region using the modified input variable to determine a new total score for the region; and

determining whether the new total score falls below the threshold.

10. The method of claim 5, further comprising:

generating a model for each region having a total score above a threshold using the input variables.

11. The method of claim 5, further comprising

identifying a total score that falls below a threshold due to at least one of insufficient data from collection sources and inconsistent calculations performed by different sources when determining the data.

12. A non-transitory computer readable medium storing computer readable instructions that when executed by a processor perform a method for determining quality of data for modeling, the method comprising:

identifying input variables to estimate a dependent variable;

determining quality metrics that quantify a sufficiency of the input variables in generating an accurate model, wherein the data is provided for a plurality of regions;

determining a quality metric score for each quality metric based on measurements for the quality metric;

determining an input variable score for each input variable based on the quality metric scores and quality metric weights for the quality metrics;

determining a total score for each region of the plurality of regions based on the input variable scores, wherein determining the total score for each region comprises;

determining categories for the input variables, wherein each category is associated with a type of input variable;

determining category weights for each category;

determining a category score for each category based on the input variable scores for the input variables associated with each category; and

determining the total score for each region based on the category scores and the category weights;

determining whether at least one of the input variables for a region of the plurality of regions is to be modified based on the total score for the region; and

upon determining that at least one of the input variables for a region of the plurality of regions is to be modified, determining whether the total score for the region is to be improved by applying a modified input variable having a different quality metric.

13. The computer readable medium of claim 12, wherein determining a quality metric score comprises:

determining a scale of values for each quality metric;

comparing the measurement for each quality metric to a range of values mapped to values within the scale; and

determining the quality metric score for each quality metric based on the comparison.

14. The computer readable medium of claim 12, wherein the categories comprise independent variables controlled by an entity, independent variables outside the entity'"'"'s control, and dependent variables that are dependent on variables in another category.

15. The computer readable medium of claim 12, wherein determining a total score for each region comprises:

determining sources for the data for the input variables;

determining source weights for each category; and

determining the total score for each region based on the source weights.

16. The computer readable medium of claim 12,

wherein the determining of whether at least one of the input variables for a region is to be modified comprises determining the at least one of the input variables for a region is to be modified upon a determination that the total score for the region falls below a threshold;

and

wherein the applying a modified input variable having a different quality metric comprises;

identifying an input variable from the input variables operable to be improved based on the quality metrics;

modifying the identified input variable;

re-scoring the region using the modified input variable to determine a new total score for the region; and

determining whether the new total score falls below the threshold.

17. The computer readable medium of claim 12, wherein the method performed by the processor includes:

identifying a total score that falls below a threshold due to at least one of insufficient data from collection sources and inconsistent calculations performed by different sources when determining the data.