Two-Tier Transaction Categorization Using Local and Global Models

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing systems face challenges in maintaining data privacy and security while using machine learning models for transaction categorization, as they often rely solely on local data which may be inaccurate and expose user information.

Innovation Solution

A two-tiered approach using a personal categorization model on a user device and a global categorization model on a central server, where transaction records are initially processed locally and sent to the central server only if the local categorization is uncertain, thereby enhancing security and data integrity.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If transaction records are processed through a global machine learning model on a central server, then categorization accuracy is improved, but data privacy and security are compromised

Engineering Contradiction:

Improvecategorization accuracyVSAvoiddata privacy exposure

Core Design Contradiction:

Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The system segments the categorization process into two distinct models: a global categorization model running on the central server that processes multiple users' data for high accuracy, and a local categorization model running on the user device that processes only local data for privacy protection. This segmentation allows each model to operate in its optimal environment while mitigating the privacy risks of centralized processing.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The local categorization model acts as an intermediary between the user's private transaction data and the global categorization system. It first processes local data and only transmits records with low confidence scores to the global model, thereby filtering and protecting sensitive information while still enabling access to the more accurate global model when necessary.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Object-affected harmful factors

If transaction records are processed only locally on user devices, then data privacy is maintained, but categorization accuracy deteriorates

Engineering Contradiction:

Improvedata privacy protectionVSAvoidcategorization accuracy

Core Design Contradiction:

Object-affected harmful factorsVSMeasurement precision

Solution Approach 1:

Instead of processing all transactions locally or all transactions globally, the system applies partial action by processing only the subset of transactions with low confidence scores through the global model. This selective approach maintains privacy for high-confidence local categorizations while improving accuracy for uncertain cases through global model intervention.

Inventive Principle:

Principle #16Partial or excessive action

3Measurement precision

If all transaction records are transmitted to the central server for processing, then categorization accuracy is improved, but system security and data exposure increase

Engineering Contradiction:

Improvecategorization accuracyVSAvoidsystem security

Core Design Contradiction:

Measurement precisionVSReliability

Solution Approach 1:

The local categorization model performs preliminary processing of transaction records before they are potentially transmitted to the central server. By pre-categorizing records and identifying only those with low confidence scores, the system reduces the volume of data transmitted to the server, thereby maintaining security while still enabling accurate categorization when needed.

Inventive Principle:

Principle #10Preliminary action

4Adaptability or versatility

If a single global model is used for all users, then model versatility is improved, but individual user data privacy is compromised

Engineering Contradiction:

Improvemodel versatilityVSAvoiduser information exposure

Core Design Contradiction:

Adaptability or versatilityVSObject-affected harmful factors

Solution Approach 1:

The system implements local quality by creating a personalized local categorization model for each user that operates on their specific device and processes their individual transaction data. This local model has quality tailored to the user's specific privacy requirements, while the global model provides versatility across multiple users. Each user's data remains localized unless confidence thresholds are not met.

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS12314442B2Multi-device, multi-model categorization system

Publication Date: 2025.05.27 INTUIT INC

AI summary

A system includes a categorization application including a personal categorization model for execution on a user device, the personal categorization model configured to assign a personal assigned category and a personal assigned categorization confidence value to transaction record. The system also includes a global categorization model executing on a central server. The global categorization model is configured to receive the transaction record from the user device, and process the transaction record through the global categorization model to obtain a global assigned category and a global assigned categorization confidence value. The system is configured to select a transaction category from the personal assigned category and the global assigned category using the personal assigned categorization confidence value and the global assigned categorization confidence value, and categorize the transaction with the transaction category.