Bank Statement Matching With Feature Vectors for Multi-Invoice Patterns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning approaches for matching items between entities, such as bank statements and invoices, are unscalable when patterns are numerous and variable, particularly in multi-match scenarios, relying on heuristics that become impractical.
Innovation Solution
A machine learning platform employing a scalable approach for multi-matching electronic documents, utilizing a machine learning model that processes feature vectors derived from aggregated pairwise feature descriptors, including encoders for different data types, to determine a probability of match between a bank statement and a super invoice, reducing the problem to a binary classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If heuristics are used to group entities for matching, then matching accuracy can be achieved for simple cases, but the approach becomes unscalable when patterns are numerous and variable
Solution Approach 1:
The patent replaces the mechanical system of manual heuristic rule maintenance with a machine learning model that automatically learns matching patterns from data. The ML model processes feature vectors derived from electronic documents and automatically determines match probabilities, eliminating the need for manual heuristic development and maintenance while adapting to numerous and variable patterns.
Solution Approach 2:
The patent transforms discrete heuristic rules into continuous feature descriptors and feature vectors. By converting categorical matching criteria into numerical features that can be processed by ML models, the system enables continuous parameter optimization and automatic adaptation to varying patterns without manual intervention.
2Productivity
If traditional matching approaches are used, then implementation is simple for single-match scenarios, but computational complexity increases significantly for multi-match scenarios
Solution Approach 1:
The patent segments the multi-match problem into independent feature extraction and classification stages. By computing feature descriptors for individual column pairs and combining them into feature vectors, the system breaks down the complex multi-match problem into manageable components that can be processed efficiently by the ML model.
Solution Approach 2:
The patent introduces feature vectors as an intermediary representation between raw document data and match decisions. This intermediate layer transforms complex document comparisons into standardized numerical vectors, enabling efficient batch processing and reducing computational complexity for multi-match scenarios.
3Measurement precision
If feature vectors with multiple modalities are used, then matching accuracy improves, but processing time and resource requirements increase
Solution Approach 1:
The patent creates a universal feature vector framework that handles multiple data modalities (strings, numbers, categories, dates) through a single standardized processing pipeline. This multi-functional approach allows the same ML model architecture to process diverse document types and fields efficiently, improving accuracy without proportionally increasing processing time.
Data Source
AI summary
Methods, systems, and computer-readable storage media for providing a set of column pairs, each column pair including a column of a bank statement table, and a column of a super invoice table, each column pair corresponding to a modality, the super invoice table including at least one row including data associated with multiple invoices, for each column pair, determining a feature descriptor based on an operator, a feature vector being provided based on feature descriptors of the set of column pairs, inputting the feature vector to a ML model that processes the feature vector to determine a probability of a match between the bank statement, and a super invoice represented by the super invoice table, and outputting a binary output representing one of a match and no match between the bank statement, and the super invoice based on the probability.