Automatic data extraction, error correction and forecasting system

• US 6,591,255 B1
• Filed: 04/05/2000
• Issued: 07/08/2003
• Est. Priority Date: 04/05/1999
• Status: Expired due to Term

• Current Assignee: Metricly, Inc. (Virtana Corp.)
• Original Assignee: Netuitive, Inc. (Virtana Corp.)
• Inventors: Jannarone, Robert J., Tatum, John T., Wilkinson, W. Clayton IV
• Primary Examiner(s): Davis, George B.
• Application Number:
  US09/543,697
• Time in Patent Office: 1,189 Days
• Field of Search: 706/25, 706/21
• US Class Current: 706/25
• CPC Class Codes: G06N 3/08 Learning methods

Abstract

A “Rapid Learner Client Service” (RLCS) system that allows a large number of end-users to obtain the benefits of a sophisticated neural-network forecasting system. Rather than purchasing or developing a forecasting system of their own, RLCS clients subscribe to a forecasting service performed by forecasting equipment located at a remote site. This allows a single highly sophisticated forecasting system to meet the forecasting needs of a large number of subscribers. This forecasting service is performed by an RLCS server that periodically and automatically accesses the subscriber'"'"'s computer to obtain a fresh set of input data. Alternatively, the subscriber'"'"'s computer may contact the RLCS server to initiate the process. This input data is then downloaded to the RLCS server, where it is checked and corrected for errors by imputing values for missing or deviant input values. The error-corrected input data is then used to compute a forecast of output values, which are downloaded to the client'"'"'s computer. The RLCS server also computes and downloads a set accuracy statistics for the client'"'"'s review.

Claims

1. A method of forecasting in a computer-based forecasting system comprising the method steps of:

receiving a set of input data from a subscriber'"'"'s computer;

downloading the input data to a remote server;

checking the input data for missing or deviant input values;

correcting for errors by imputing values for the missing or the deviant input values;

computing a forecast of output values based on the set of the input data; and

downloading the output values to the subscriber'"'"'s computer.

2. The method of claim 1, wherein the checking step comprises the method steps of:

computing deviance values from the set of input data;

determining if an override option and error correction and detection option have been enabled;

if the override option and the error correction and detection option have been enabled, determining errors in the deviance values;

correcting errors in the deviance values to create a corrected set of input data; and

performing imputation learning using the corrected set of input data.

3. The method of claim 2, wherein the correcting step comprises the method steps of:

assessing the deviance values;

identifying a deviant deviance value which exceeds a predefined threshold;

excluding the deviant deviance value and setting the deviant deviance value to a non-viable deviant value;

re-computing the deviance values; and

repeating the accessing, identifying and excluding steps until all of the deviance values are below the predefined threshold.

4. The method of claim 2, wherein the step of performing imputation learning includes the steps of:

updating imputing function means;

imputing function variances;

imputing function connection weights; and

updating error variances used to compute the deviance values.

5. The method of claim 1, wherein:

the checking step further comprises the steps of;

identifying a missing value from the set of input data;

disregarding the missing value and setting the missing value as a non-viable input;

imputing a value for such missing value;

correcting the missing value with the imputed value; and

performing imputation learning using the corrected missing value and the set of input data.

6. The method of claim 5, wherein the step of performing imputation learning includes the steps of:

updating error variances for only viable input data; and

updating kernel learned parameters including feature means and connection weights.

7. The method of claim 6, wherein:

the connection weights define elements of an inverse covariance matrix; and

the step of imputing includes the steps of;

automatically updating the connection weights in a covariance matrix corresponding to the inverse covariance matrix; and

inverting the updated covariance matrix.

8. The method of claim 6, wherein:

the connection specifications include connection weights defining elements of an inverse covariance matrix; and

the method step of imputing comprises the step of;

automatically updating the connection weights of the inverted covariance matrix.

9. A method for detecting and selectively correcting errors in a set of input data for use in a computer forecasting system comprising the steps of:

receiving a value for each input data of the set of input data;

identifying a missing value from the set of input data;

disregarding the missing value and setting such missing value as a non-viable input;

computing deviance values from the set of input data;

correcting the non-viable inputs;

performing imputation learning using the corrected non-viable inputs and the set of input data; and

computing a forecast based on the set of input data.

10. The method of claim 9, further comprising the method steps of:

determining if an override option and error correction and detection option have been enabled; and

if the override option and the error correction detection option have been enabled performing,

assessing deviance values,

identifying a deviant deviance value which exceeds a predefined threshold,

excluding the deviant deviance value and setting the deviant deviance value to a non-viable deviant value;

re-computing the deviance values, and

repeating the assessing, identifying and excluding steps until all of the deviance values are below the predefined threshold.

11. The method of claim 10, further comprising the method steps of:

correcting the non-viable inputs; and

performing imputation learning using the corrected non-viable inputs and the set of input data.

12. A computer-based forecasting system comprising:

means for receiving a set of input data from a subscriber'"'"'s computer;

means for downloading the input data to a remote server;

means for checking the input data for missing or deviant input values;

means for correcting for errors by imputing values for the missing or the deviant input values;

means computing a forecast of output values based on the set of input data; and

means for downloading the output values to the subscriber'"'"'s computer.