Unified ML Model for Segmented Risk Prediction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing machine learning models for segment-based risk assessment and outcome prediction face inconsistencies in prediction scales across different segments, leading to delayed and distorted results due to post-hoc alignment processes.

Innovation Solution

A unified machine learning model is generated by stacking and training multiple segment models, eliminating the need for post-hoc alignment by aligning values through hidden and output layers, thereby improving prediction accuracy and speed.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If different models are built for different entity segments, then prediction accuracy for each segment is improved, but prediction consistency across segments deteriorates due to different scales

Engineering Contradiction:
Improveprediction accuracyVSAvoidprediction consistency
Core Design Contradiction:
Measurement precisionVSStability of the object's composition

Solution Approach 1:

The patent merges multiple segment-specific models into a single unified model that processes all entity segments simultaneously. This unified model maintains the specialized knowledge for each segment while ensuring consistent prediction scales across all segments, resolving the contradiction between segment-specific accuracy and cross-segment consistency.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The unified model is designed to handle multiple entity segments with different risk profiles using a single universal architecture. This multi-functional approach allows the model to maintain specialized prediction capabilities for each segment type while operating on a consistent scale, eliminating the need for separate models.

Inventive Principle:
Principle #6Universality (Multi-functionality)

2Stability of the object's composition

If post-hoc alignment is introduced to align prediction outputs from different models, then prediction consistency across segments is improved, but prediction speed deteriorates and prediction accuracy deteriorates due to introduced distortions

Engineering Contradiction:
Improveprediction consistencyVSAvoidprediction speed
Core Design Contradiction:
Stability of the object's compositionVSProductivity

Solution Approach 1:

The unified model performs alignment of prediction scales during the model design and training phase rather than as a post-processing step. This preliminary action ensures that predictions from different entity segments are generated on consistent scales from the outset, eliminating the need for subsequent alignment operations that would slow down prediction.

Inventive Principle:
Principle #10Preliminary action

3Stability of the object's composition

If post-hoc alignment is introduced to align prediction outputs from different models, then prediction consistency across segments is improved, but prediction accuracy deteriorates due to introduced distortions

Engineering Contradiction:
Improveprediction consistencyVSAvoidprediction accuracy
Core Design Contradiction:
Stability of the object's compositionVSMeasurement precision

Solution Approach 1:

The unified model aligns prediction scales during training through appropriate loss functions and normalization techniques applied to the training data. This preliminary alignment prevents distortion introduction that would occur with post-hoc alignment, maintaining both consistency and accuracy.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent replaces the mechanical post-hoc alignment process with a statistical learning approach where the unified model learns appropriate scaling and normalization during training. This substitution eliminates the need for separate alignment operations that introduce distortions, preserving prediction accuracy while achieving consistency.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentEP4202771A1Unified explainable machine learning for segmented risk assessment
Publication Date: 2023.06.28 EQUIFAX INC
  • EP4202771A1 patent drawingFigure 1
  • EP4202771A1 patent drawingFigure 2
  • EP4202771A1 patent drawingFigure 3

AI summary

Various aspects involve unified explainable machine learning for segmented risk assessment. For example, a computing device can determine, using a unified model built from segment models, a risk indicator for a target entity from predictor variables associated with the target entity. The target entity belongs to one of a plurality of entity segments each associated with a segment model of the segment models. The unified model is generated by: accessing training samples for the entity segments; training the segment models using respective training samples for the entity segments; constructing the unified risk prediction model by stacking the trained segment models; and training the unified risk prediction model using the training samples for the entity segments. The computing device can transmit, to a remote computing device, a responsive message including at least the risk indicator for use in controlling access of the target entity to an interactive computing environment.