Predictive model augmentation by variable transformation

• US 20050234763A1
• Filed: 04/16/2004
• Published: 10/20/2005
• Est. Priority Date: 04/16/2004
• Status: Active Grant

• Current Assignee: Longhorn Automotive Group LLC (Alpha Alpha Intellectual Partners LLC)
• Original Assignee: Fortelligent LLC
• Inventors: Pinto, Stephen K., Jacobs, Marc, Mansfield, Richard, Rubin, Donald
• Granted Patent: US 7,730,003 B2
• Time in Patent Office: Days
• US Class Current: 705/10
• CPC Class Codes: G06Q 10/04 Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem" market predictions or forecasting for commercial activities G06Q30/0202
• CPC Class Codes: G06Q 30/0201 Market modelling; Market analysis; Collecting market data

Abstract

Models are generated using a variety of tools and features of a model generation platform. For example, in connection with a project in which a user generates a predictive model based on historical data about a system being modeled, the user is provided through a graphical user interface a structured sequence of model generation activities to be followed, the sequence including dimension reduction, model generation, model process validation, and model re-generation. Historical multi-dimensional data is received representing multiple source variables to be used as an input to a predictive model of a commercial system and applying transformations to the data that are selected based on the strength of measurement represented by a variable; variables are transformed into new more predictive variables, including the Bayesian renormalization of sparsely sampled variable and including the imputation of missing values for categorical or continuous variables.

Claims

1. A machine-based method comprising

receiving historical multi-dimensional data representing multiple source variables to be used as an input to a predictive model of a commercial system,

applying transformations to the source variables that are selected to increase predictive power, and

applying transformations to the data that are selected based on the strength of measurement represented by a variable.

2. The method of claim 1 in which the strength of measurement comprises at least one of nominal, ordinal, and interval.

3. The method of claim 1 in which the strength of a measurement is represented in stored metadata associated with the data.

4. The method of claim 1 also including

displaying to a user a representation of a response function of a target variable against untransformed, transformed, and target variables associated with the data.

5. The method of claim 1 also including

persistently storing both the source variables and related transformed versions of the source variables.

6. A machine-based method comprising

receiving historical multi-dimensional data representing multiple source variables to be used as an input to a predictive model of a commercial system,

adjusting unstable values of the variables to reduce inaccurate associations between predictor variables and target variables.

7. The method of claim 6 in which the adjustment of the unstable values comprises Bayesian renormalization.

8. A machine-based method comprising

in connection with a project in which a user generates a predictive model based on historical data about a system being modeled, automatically imputing missing values for continuous variables associated with the data.

9. The method of claim 8 in which the user is enabled to invoke the automatic imputing as part of a user interface feature that displays information about variables for which values are missing.

10. The method of claim 9 in which the automatic imputing is invoked based on the variable or type of variable.

11. The method of claim 9 in which the variables for which missing values are imputed may be used in the model or may be transformed for use in the model.