Agnostic Data Structure for Predictive Model Debriefing

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The existing debriefing process for predictive models is inefficient as it requires a running instance of analytic software, consuming significant system resources and time, and is difficult to interact with, especially when generating debrief information for different model types and user profiles.

Innovation Solution

An agnostic data structure is introduced that packages debriefing information in a generic schema, allowing for efficient querying and generation of debriefs without the need for a running software instance, using a multi-layered format with a generic storage layer and semantic layer for model performance and relationship representation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If a running instance of analytic software is loaded into the testing environment to generate a debrief, then complete model performance information can be obtained, but significant system resources are occupied and additional time is required during each iteration

Engineering Contradiction:

Improvecompleteness of model performance informationVSAvoiddebrief generation speed

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The patent extracts and stores model performance information into a debrief data structure during the training process, before debrief generation is needed. This preliminary capture of performance metrics eliminates the need to load the running analytic software instance solely for debrief generation, thereby reducing system resource occupation and accelerating the debrief generation process while maintaining complete model performance information

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent extracts specific model performance information from the running analytic software instance and stores it in a separate debrief data structure. By taking out only the necessary performance metrics and storing them independently, the system can generate debriefs without loading the entire analytic software, thus improving productivity while preserving the reliability of performance data

Inventive Principle:

Principle #2Taking out (Extraction)

2Reliability

If a running instance of analytic software is loaded into the testing environment to generate a debrief, then accurate model information can be captured, but the process becomes more complex and resource-intensive

Engineering Contradiction:

Improveaccuracy of model informationVSAvoiddebrief generation process complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The patent extracts model performance information from the complex running analytic software instance and stores it in a simplified debrief data structure. This extraction process captures accurate model information while separating it from the complexity of the running software, thereby reducing the complexity of the debrief generation process

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent creates a copy of the essential model performance information in a debrief data structure that can be generated without the running software instance. This copy contains accurate model information in a simplified format that is easier to generate and manage, reducing process complexity while maintaining information accuracy

Inventive Principle:

Principle #26Copying

3Adaptability or versatility

If multiple software instances are loaded for different model types and user profiles, then comprehensive debriefing coverage is achieved, but system resource consumption increases significantly

Engineering Contradiction:

Improvedebriefing coverage across model types and user profilesVSAvoidsystem resource consumption

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The patent creates a universal debrief data structure that can accommodate multiple model types and user profiles through a common interface and standardized format. This universal structure allows the system to generate comprehensive debriefing coverage for different model types and user profiles without loading multiple separate software instances, thereby reducing system resource consumption while maintaining adaptability

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent performs preliminary extraction and storage of model performance information in a universal debrief data structure that is designed to handle multiple model types and user profiles. This preliminary action enables the system to serve multiple purposes with a single data structure, eliminating the need for multiple software instances and reducing overall system resource consumption

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS11468339B2Agnostic data structure for debriefing predictive software

Publication Date: 2022.10.11 SAP SE

AI summary

Provided are systems and methods for generating an agnostic data structure that stores debriefing information for a predictive model. In one example, the method may include receiving training data of a predictive program having a model type from among a plurality of different model types, identifying values of generic debriefing information from the training data which is generic among the different model types and values of semantic debriefing information from the training data which is unique to the model type of the received predictive program from among the plurality of different model types, extracting the values of the generic debriefing information and the values of the semantic debriefing information, and storing the values of the generic debriefing information and the semantic debriefing information within an agnostic debriefing data structure.