Fraud Score Normalization via Intermediary Mediator

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing fraud detection and identity verification systems using machine learning models face challenges in maintaining consistent fraud scoring interpretations due to biases and variability in input data, leading to inconsistent results over time.

Innovation Solution

A system and method for iteratively transforming scores data by sorting, aggregating, and mapping them into a normalized representation, independent of the machine learning model type, to achieve uniformity and ease of interpretation across different models and transactions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If machine learning models are used to generate fraud scores, then fraud detection capability is improved, but scoring consistency and interpretability deteriorate due to model-type variability and input bias

Engineering Contradiction:
Improvefraud detection capabilityVSAvoidscoring consistency
Core Design Contradiction:
ReliabilityVSMeasurement precision

Solution Approach 1:

The patent introduces a normalization layer as an intermediary between machine learning models and fraud scoring interpretation. This normalization component standardizes scores from different model types and time periods, acting as a mediator that enables consistent interpretation without requiring changes to the underlying models themselves.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent transforms fraud scores through parameter changes by applying normalization functions that adjust score distributions. This involves changing the statistical parameters (mean, variance, distribution shape) of the scores to achieve consistency across different models and time periods, while preserving the relative ranking and predictive power.

Inventive Principle:
Principle #35Parameter changes

2Adaptability or versatility

If machine learning models with varying input data are used, then fraud detection versatility is improved, but scoring uniformity deteriorates

Engineering Contradiction:
Improvefraud detection versatilityVSAvoidscoring uniformity
Core Design Contradiction:
Adaptability or versatilityVSStability of the object's composition

Solution Approach 1:

The patent creates a universal normalization framework that can handle scores from multiple different machine learning model types and input configurations. This universal approach allows the system to accommodate diverse model inputs while producing uniformly interpretable outputs, enabling one normalization mechanism to serve multiple model types.

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

Solution Approach 2:

The patent applies equipotentiality by creating a common reference frame for all fraud scores through normalization. This establishes an equipotential state where scores from different models with different inputs are adjusted to the same distributional level, making them comparable and uniformly interpretable despite their diverse origins.

Inventive Principle:
Principle #12Equipotentiality

3Productivity

If fraud scores are interpreted directly from machine learning models, then processing speed is improved, but interpretation accuracy deteriorates due to bias and variability

Engineering Contradiction:
Improveprocessing speedVSAvoidinterpretation accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies preliminary action by performing normalization of fraud scores before they are interpreted or used for decision-making. This pre-processing step ensures that scores are standardized and free from model-specific biases before reaching the interpretation stage, improving accuracy without adding significant processing time.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent implements feedback mechanisms where normalization parameters are continuously refined based on observed score distributions and their interpretability. This feedback loop allows the system to learn from interpretation outcomes and adjust normalization parameters to improve both accuracy and consistency over time.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP4075363A1Self learning machine learning scores adjustment via normalization thereof
Publication Date: 2022.10.19 SOCURE INC
  • EP4075363A1 patent drawingFigure 1
  • EP4075363A1 patent drawingFigure 2
  • EP4075363A1 patent drawingFigure 3

AI summary

Provided are a system and methodology for iteratively transforming data as between multiple sets thereof. Doing so, via normalization of the data, enables uniformity of interpretation and presentation of the data no matter the machine learning model that produced the data.