Model integration tool
AssigneeEquifax Inc.
InventorsRajesh Indurthivenkata, Lalithadevi Venkataramani, Aparna Somaka, Xingjun Zhang, Matthew Turner, Bhawana Koshyari, Vijay Nagarajan, James Reid, Nandita Thakur
Life of this patent
Estimating the term…
Abstract
Certain aspects involve building and debugging models for generating source code executed on data-processing platforms. A target data-processing platform is identified that requires bin ranges for modeling coefficients and reason codes for input attributes. A processor outputs source code, which is generated from a modeling code, in a programming language used by the target data-processing platform.
Claims17
A model integration system comprising:
- a processor communicatively coupled via a data network to a client computing system;
- and
- a non-transitory computer-readable medium storing program code that provides a model integration tool and that is executable by the processor;
- wherein the processor is configured for executing the program code to perform operations comprising:
- receiving an electronic data-processing model that generates an analytical output by applying a function to one or more input attributes weighted with one or more modeling coefficients, respectively,
- identifying a target data-processing platform that requires one or more bin ranges associated with the one or more modeling coefficients and one or more reason codes associated with the one or more input attributes, wherein each bin range identifies a range of input attribute values for which a respective modeling coefficient is valid, wherein each reason code identifies a respective reason code output indicating a respective impact of a respective weighted input attribute on the analytical output,
- providing a user interface having interface elements for specifying bin range information identifying the one or more bin ranges and reason code information identifying the one or more reason codes, the user interface having one or more additional interface elements for uploading, to the non-transitory computer-readable medium or an additional non-transitory computer-readable medium accessible to the model integration tool, an audit file having an expected analytical output and an expected reason code,
- identifying the one or more bin ranges and the one or more reason codes based on user input received via the user interface and indicating the bin range information and the reason code information,
- generating Predictive Model Markup Language (PMML) code that implements the electronic data-processing model with the one or more bin ranges and the one or more reason codes,
- verifying, from a test of the PMML code and prior to outputting executable source code in a programming language used by the target data-processing platform, that a test analytical output matches the expected analytical output and that a test reason code matches the expected reason code, wherein verifying that the test analytical output matches the expected analytical output and the test reason code matches the expected reason code comprises:
- (i) accessing the audit file from the non-transitory computer-readable medium or the additional non-transitory computer-readable medium, and
- (ii) comparing the expected analytical output and the expected reason code stored in the audit file with the test analytical output and the test reason code, and
- outputting the executable source code in the programming language used by the target data-processing platform, wherein the executable source code is generated from the PMML code and implements the electronic data-processing model with the one or more bin ranges and the one or more reason codes.
A method comprising:
- receiving, by a model integration tool executed by a processing device, an electronic data-processing model that generates an analytical output by applying a function to one or more input attributes weighted with one or more modeling coefficients, respectively, wherein the received electronic data-processing model lacks bin range information associated with the one or more modeling coefficients and reason code information associated with the one or more input attributes;
- providing, from the model integration tool to a client computing system, a user interface having interface elements for specifying the bin range information and the reason code information, the user interface having additional interface elements for uploading, to a non-transitory computer-readable medium accessible to the model integration tool, an audit file having an expected analytical output and an expected reason code;
- identifying, by the model integration tool and based on user input received via the interface elements, one or more bin ranges for the one or more modeling coefficients and one or more reason codes for the one or more input attributes, wherein each bin range identifies a range of input attribute values for which a respective modeling coefficient is valid, wherein each reason code identifies a respective reason code output indicating a respective impact of a respective weighted input attribute on the analytical output;
- generating, by the model integration tool, modeling code that specifies the electronic data-processing model with the one or more bin ranges and the one or more reason codes;
- verifying, with the model integration tool, that a test of the modeling code results in a test analytical output matching the expected analytical output and a test reason code matching the expected reason code, the verification comprising accessing the audit file from the non-transitory computer-readable medium and comparing the expected analytical output and the expected reason code stored in the audit file with the test analytical output and the test reason code;
- and
- generating, by the model integration tool and subsequent to the verification, an executable source code that implements the electronic data-processing model on a data-processing platform that is accessible by the client computing system, wherein the executable source code is generated from the modeling code.
A non-transitory computer-readable medium having program code of a model integration tool that is executable by a processing device to cause a computing device to perform operations, the operations comprising:
- receiving an electronic data-processing model that generates an analytical output by applying a function to one or more input attributes weighted with one or more modeling coefficients, respectively;
- providing, to a client computing system, a user interface having interface elements for specifying bin range information associated with the one or more modeling coefficients and reason code information associated with the one or more input attributes, the user interface having additional interface elements for uploading, to a memory device that is at least one of the non-transitory computer-readable medium or an additional non-transitory computer-readable medium accessible to the model integration tool, an audit file having an expected analytical output and an expected reason code;
- identifying, based on input received via the interface elements, one or more bin ranges for the one or more modeling coefficients and one or more reason codes for the one or more input attributes, wherein each bin range identifies a range of input attribute values for which a respective modeling coefficient is valid, wherein each reason code identifies a respective reason code output indicating a respective impact of a respective weighted input attribute on the analytical output;
- generating Predictive Model Markup Language (PMML) code that specifies the electronic data-processing model with the one or more bin ranges and the one or more reason codes;
- verifying that a test of the PMML code results in a test analytical output matching the expected analytical output and a test reason code matching the expected reason code, the verification comprising accessing the audit file from the memory device and comparing the expected analytical output and the expected reason code stored in the audit file with the test analytical output and the test reason code;
- and
- generating, subsequent to the verification, an executable source code that implements the electronic data-processing model on a data-processing platform that is accessible by the client computing system, wherein the executable source code is generated from the PMML code.
Dependencies are inferred from the claim text.