Browse Prior Art Database

A Method and System for Verifying Predictive Model Outputs

IP.com Disclosure Number: IPCOM000194685D
Publication Date: 2010-Apr-07
Document File: 1 page(s) / 200K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a method and system for verifying predictive model outputs. The method and system enables a user to automatically identify discrepancies between the predictive model outputs and one or more of domain constraints as specified by the user and aggregate measures computed on a training data.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 99% of the total text.

Page 1 of 1

A Method and System for Verifying Predictive Model Outputs

A Method and System for Verifying Predictive Model OutputsA Method and System for Verifying Predictive Model Outputs

                                             . The method and system enables a user to automatically identify discrepancies between the predictive model outputs and one or more of domain constraints as specified by the user and aggregate measures computed on a training data .

The figure illustrates the system for verifying predictive model outputs .

Figure

As shown in the figure, aggregate measures are computed over the training data used

by the predictive model. The aggregate measures are stored in an aggregate measured DB. The system also includes a domain constraints DB that stores domain constraints as specified by a user.

A validation engine takes into account one or more

of the domain constraints and the aggregate measures to match it against the

predictive model outputs on test data and identify discrepancies .

Thereafter, the discrepancies are displayed to a user in terms of list of the aggregate measures and the domain constraints which are not satisfied by the predictive model output. The discrepancies are shown in a manner that enables easy assimilation by a non-technical user and enables him to make a decision on whether the predictive model output may be rejected or approved for further downstream processing .

A method and system is disclosed for verifying predictive model outputs

1

[This page contains 1 picture or other...