ML Binary Classifiers for Invoice Super-Invoice Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing approaches for matching bank statements to multiple invoices are inefficient and non-scalable due to the combinatorial complexity of the subset sum problem, leading to impractical brute-force search methods in real-world applications.
Innovation Solution
A machine learning platform utilizing binary classifiers, including support vector machines and deep learning models, processes invoices to form cliques based on feature comparisons and aggregate features, enabling efficient matching of bank statements to super-invoices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional brute-force search methods are used to match bank statements to multiple invoices, then completeness of matching is improved, but computational complexity and processing time increase exponentially
Solution Approach 1:
The patent segments the complex multi-match problem into two simpler binary classification problems: (1) determining whether a bank statement matches a single invoice, and (2) determining whether a bank statement matches a set of invoices. This segmentation transforms the exponentially complex combinatorial optimization problem into manageable binary decisions that can be processed efficiently by machine learning models.
Solution Approach 2:
The patent introduces an intermediary representation called 'super-invoice' which aggregates multiple individual invoices into a single composite entity. This super-invoice serves as a mediator between the bank statement and individual invoices, allowing the system to first match the bank statement to the super-invoice set, then decompose into individual invoice matches, thereby avoiding direct brute-force comparison of the bank statement against all possible invoice combinations.
2Productivity
If heuristics are used to group matched entities, then processing speed is improved, but scalability deteriorates when patterns become numerous and variable
Solution Approach 1:
The patent transforms the matching problem from using fixed heuristic rules to using learnable parameters through machine learning. The binary classifiers are trained on historical matching data, allowing the system to automatically adapt to varying patterns and domains. This parameter transformation enables the system to maintain high processing speed while gaining scalability to handle numerous and variable matching patterns across different domains.
Solution Approach 2:
The patent replaces the mechanical heuristic-based grouping system with a data-driven machine learning approach. Instead of manually designed rules that require reconfiguration for different patterns, the system uses trained binary classifiers that automatically learn optimal matching strategies from data, providing both speed and adaptability to variable patterns.
3Adaptability or versatility
If binary classifiers are used to determine matching, then scalability is improved, but measurement precision requirements increase
Solution Approach 1:
The patent performs preliminary feature extraction and aggregation before the binary classification decision. By pre-computing aggregate features of invoice sets and preparing structured input representations, the system reduces the computational burden during the actual classification step. This preliminary preparation enables the binary classifiers to operate efficiently with high precision requirements without sacrificing scalability.
Data Source
AI summary
Methods, systems, and computer-readable storage media for providing, by a machine learning (ML) platform, a first binary classifier, processing, by the first binary classifier a super-set of invoices to provide a plurality of sets of invoices based on matching pairs of invoices in the super-set of invoices, providing, by the ML platform, a second binary classifier, processing, by the second binary classifier, a bank statement and the plurality of sets of invoices to define two or more super-invoices based on aggregate features of invoices in the plurality of sets of invoices, and match the bank statement to a super-invoice of the two or more super-invoices, and outputting a match of the bank statement to the super-invoice.