Transaction Signature Distance for Cross-Institution Categorization

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current techniques for transaction communication and categorization in financial institutions are prone to errors and are difficult to apply across multiple institutions due to variations in transaction descriptors.

Innovation Solution

A method and system for transaction analysis that extracts statistically correlated character groups from transaction identifiers, generates transaction signatures using inverse document frequency values, and computes distances between signatures to determine category membership.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional transaction categorization methods are used, then transactions can be grouped by category, but the accuracy is reduced due to variations in transaction descriptors across different financial institutions

Engineering Contradiction:

Improvetransaction categorization accuracyVSAvoidapplicability across multiple financial institutions

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent transforms transaction descriptors from variable-length text with inconsistent formatting into fixed-length numeric vectors through embedding models. This parameter transformation enables consistent mathematical comparison across different financial institutions' descriptors, resolving the contradiction between categorization accuracy and cross-institution adaptability

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent introduces cosine similarity as an intermediary metric between transaction descriptors from different institutions. This intermediary enables standardized comparison by measuring the angular relationship between embedding vectors, allowing accurate categorization across institutions with varying descriptor formats

Inventive Principle:

Principle #24Intermediary (Mediator)

2Adaptability or versatility

If standardized transaction analysis methods are implemented, then cross-institution compatibility improves, but the complexity of processing and analyzing transaction data increases

Engineering Contradiction:

Improvecross-institution compatibilityVSAvoiddata processing complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent performs preliminary embedding transformations on transaction descriptors before categorization analysis. By pre-converting descriptors into standardized vector representations, the system reduces the complexity of subsequent comparisons across institutions, as the heavy lifting of normalization is already completed

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent replaces traditional text-based transaction analysis mechanisms with vector-based mathematical operations. This substitution enables cross-institution compatibility through standardized linear algebra operations (cosine similarity calculations) rather than complex text parsing and pattern matching

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS10657525B2Method and apparatus for determining expense category distance between transactions via transaction signatures

Publication Date: 2020.05.19 KASISTO INC

AI summary

A method for transaction analysis comprises extracting a plurality of character groups from a plurality of transaction identifiers, determining an inverse document frequency value for each of the plurality of character groups, generating, for a first transaction and a second transaction, respectively, a first transaction signature and a second transaction signature corresponding to the plurality of character groups, and computing a distance between the first and second transaction signatures.