Prediction Explanation Using Influential Variable Resampling

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing machine learning models lack effective methods to provide deterministic, computationally efficient, and industry-acceptable explanations for decisions, particularly in regulated industries, leading to sub-optimal decision-making and restricted fraud reduction or credit provision.

Innovation Solution

An explanation procedure that identifies the most influential input variables on a prediction by re-sampling the prediction surface around the prediction, using metrics adjusted by prior probabilities to account for categorical and continuous variables, and generating justifiable explanations through graphical and textual outputs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If LIME methodology is used to provide explanations for machine learning model decisions, then explanations can be generated for complex models, but the method becomes computationally expensive and produces non-deterministic results

Engineering Contradiction:

Improveability to explain different machine learning modelsVSAvoidcomputational efficiency

Core Design Contradiction:

Adaptability or versatilityVSProductivity

Solution Approach 1:

The system pre-computes and stores feature importance metrics and prediction surface characteristics during model training. During real-time explanation generation, these pre-computed metrics are retrieved and applied directly to new predictions, eliminating the need for expensive post-hoc computation while maintaining explanation accuracy and determinism.

Inventive Principle:

Principle #10Preliminary action

2Adaptability or versatility

If LIME methodology is used to generate explanations, then explanations can be provided for various model types, but the explanations do not directly match regulatory industry requirements

Engineering Contradiction:

Improvecompatibility with different machine learning modelsVSAvoidalignment with regulatory requirements

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The system transforms explanation outputs by adjusting parameters such as feature ranking, threshold settings, and presentation formats to match specific regulatory requirements. The explanation generator can re-rank features based on regulatory priorities and format outputs according to industry standards while maintaining the underlying model-agnostic explanation methodology.

Inventive Principle:

Principle #35Parameter changes

3Reliability

If scorecard technology is used to generate models and explanations, then explanations can be produced in a deterministic manner, but the models cannot handle complex data types and relationships used in fraud determinations

Engineering Contradiction:

Improvedeterminism of explanationsVSAvoidability to model complex data relationships

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The system segments the explanation generation process into two independent components: a flexible machine learning model that handles complex data relationships and fraud detection, and a separate explanation generator that provides deterministic, regulatory-compliant explanations. This segmentation allows each component to optimize for its specific function without compromising the other.

Inventive Principle:

Principle #1Segmentation

4Measurement precision

If advanced machine learning models are used to optimize decision performance, then decision accuracy improves, but the ability to generate industry-acceptable explanations deteriorates

Engineering Contradiction:

Improvedecision accuracyVSAvoidease of generating acceptable explanations

Core Design Contradiction:

Measurement precisionVSEase of operation

Solution Approach 1:

The system introduces an explanation generator as an intermediary component between the advanced machine learning model and the decision-making process. This intermediary translates complex model predictions into simple, actionable explanations that satisfy regulatory requirements while maintaining the high accuracy of the underlying machine learning model.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS20260023990A1Methods of explaining an individual predictions made by predictive processes and/or predictive models

Publication Date: 2026.01.22 SYNCHRONY BANK

AI summary

A computer-implemented method that includes obtaining a plurality of values each corresponding to one of a plurality of variables. The plurality of variables include variables of interest. The method includes obtaining a prediction for the values from a model, determining metric(s) for each of the variables of interest, and determining one or more of the variables of interest to be one or more influential variables based on the metric(s) determined for each of the variables of interest. The variables include one or more non-influential variables that is/are different from the influential variable(s). The influential variable(s) has/have a greater influence on the prediction than the non-influential variable(s). The method also includes displaying in a graphical user interface or printing in a report an explanation identifying the influential variable(s) and/or a justification of the determination that the influential variable(s) has/have a greater influence on the prediction than the non-influential variable(s).