Standardized Modeling Suite

• US 20120239375A1
• Filed: 03/17/2011
• Published: 09/20/2012
• Est. Priority Date: 03/17/2011
• Status: Active Grant

• Current Assignee: Bank of America Corp.
• Original Assignee: Bank of America Corp.
• Inventors: Laxmanan, Kasilingam Basker, Chen, Yudong, Parise, Gerald F., Ware, Lewis Ray
• Granted Patent: US 8,583,408 B2
• Time in Patent Office: Days
• US Class Current: 703/22
• CPC Class Codes: G06Q 10/06 Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling

Abstract

An enhanced modeling tool associated with an entity may facilitate end-to-end modeling of problems in any application space. The enhanced modeling tool may collect modeling data from a variety of sources, check the collected data, find the best predictor variables for a given target variable, estimate the model, implement the model, and validate the model. The output of each of these steps may be in a standardized format to allow other steps to directly incorporate the output. An additional feature of the system may include a reporting capability that generates supporting documents related to model governance and risk compliance.

Claims

1. A non-transitory computer-readable storage medium having computer-executable program instructions stored thereon that when executed by a processor, cause the processor to perform steps comprising:

(i) collecting data from a plurality of sources, wherein the collected data includes data related to a plurality of independent variables;

(ii) storing the collected data;

(iii) performing a plurality of checks on the collected data, wherein the plurality of checks include handling missing values, eliminating outlier values, and eliminating highly correlated independent variables in the collected data;

(iv) analyzing the checked data to determine one or more of the plurality of independent variables to be used in predicting a target variable;

(v) generating a model to estimate the target variable using the determined independent variables; and

(vi) generating a plurality of graphs related to governance and risk compliance of the model, wherein the plurality of graphs include a variance inflation factor.

2. The computer-readable storage medium of claim 1, wherein the plurality of sources includes customer account information and transaction information.

3. The computer-readable storage medium of claim 1, wherein the data include data relevant to credit card, small business, and deposit lines of business.

4. The computer-readable storage medium of claim 1, wherein the plurality of independent variables includes derived variables derived from at least two of the plurality of independent variables.

5. The computer-readable storage medium of claim 1, wherein the processor further performs:

displaying an inventory screen with a listing of all of the plurality of independent variables that are available for modeling purposes.

6. The computer-readable storage medium of claim 5, wherein the processor further performs:

displaying derivation logic for derived variables that are derived from multiple independent variables on the inventory screen.

7. The computer-readable storage medium of claim 1, wherein the missing values are handled by filling in the missing values with a median value for one of the plurality of independent variables.

8. The computer-readable storage medium of claim 1, wherein the outlier values are defined as values that are below a 0.5 percentile or above a 99.5 percentile for each of the plurality of independent variables.

9. The computer-readable storage medium of claim 1, wherein the plurality of checks further include eliminating one of the plurality of independent variables from inclusion in the model when more than 35% of values for the one of the independent variables are missing.

10. The computer-readable storage medium of claim 1, wherein the processor further performs:

ranking the plurality of independent variables for predictive power by boot screening the plurality of independent variables.

11. The computer-readable storage medium of claim 1, wherein the determining further comprises:

determining interaction variables from interaction of at least two of the plurality of independent variables.

12. The computer-readable storage medium of claim 1, wherein the processor further performs:

segmenting the plurality of independent variables to generate two separate models from the data.

13. The computer-readable storage medium of claim 1, wherein the determining further comprises:

determining which functional form of the plurality of independent variables should be used in the model, wherein the functional forms of the plurality of independent variables are created through non-linear transformations of the plurality of independent variables.

14. A computer-assisted method comprising:

collecting data from a plurality of data sources, wherein the collected data includes data related to a plurality of independent variables;

storing the collected data in a datastore associated with the computer;

using a processor associated with the computer, performing a plurality of checks on the collected data, wherein the plurality of checks include replacing missing values in the collected data with a median value of one of the plurality of independent variables, removing outlier values that are below a 0.5 percentile or above a 99.5 percentile for each of the plurality of independent variables in the collected data, and normalizing each of the plurality of independent variables;

using the processor, analyzing the checked data to identify a subset of the plurality of independent variables for inclusion in a model for a dependent variable, wherein the analysis includes transforming categorical variables into numerical variables; and

using the processor, generating the model using the identified subset of independent variables.

15. The computer-assisted method of claim 14, further comprising:

generating a plurality of graphs related to governance and risk compliance of the model.

16. The computer-assisted method of claim 14, wherein the dependent variable is either a binary variable or a continuous variable.

17. The computer-assisted method of claim 14, wherein the model is chosen from the group consisting of:

a linear regression, a logistic regression, and a survival analysis.

18. An apparatus comprising:

a data collection module configured to collect data from a plurality of sources, wherein the collected data includes data related to a plurality of independent variables;

a data checking module configured to perform a plurality of checks on the collected data, wherein the plurality of checks include filling in missing values in the collected data with a median value of one of the plurality of independent variables, removing outlier values, removing highly correlated independent variables, and normalizing at least one of the plurality of independent variables;

a data analysis module configured to identify a first subset of the plurality of independent variables for inclusion in a model for a dependent variable by correlating at least one of the plurality of independent variables with the dependent variable through a plurality of statistics, wherein the data analysis module is further configure to transform categorical variables into numerical variables;

a data reduction module configured to identify a second subset of strongest independent variables from the first subset by boot screening the first subset of the plurality of independent variables; and

a modeling module configured to estimate the model based on the identified second subset of strongest independent variables.

19. The apparatus of claim 18, further comprising:

a data segmentation module configured to segment the plurality of independent variables for use in at least two separate models.

20. The apparatus of claim 18, further comprising:

a data transformation module configured to find a most predictive functional form of each of the plurality of independent variables.