Merchant Embedding Clustering for Accurate Store Data Merging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Merchant stores are often incorrectly identified in transaction data, leading to confusion and inaccurate analysis of spending habits and fraudulent transactions due to inconsistent naming conventions.
Innovation Solution
A system utilizing pre-trained word embeddings refined by machine-learning models to identify and aggregate merchant stores by fine-tuning embeddings with merchant-specific data, clustering similar embeddings, and normalizing cluster names to accurately merge them into a single merchant identifier.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If merchant stores are identified using different naming conventions in transaction data, then the system can capture diverse merchant representations, but the accuracy of merchant identification deteriorates causing incorrect identification of merchant stores
Solution Approach 1:
The patent merges multiple merchant identifier clusters into a single unified merchant entity by comparing cluster names and determining similarity. When clusters represent the same merchant (e.g., 'Walmart Store' and 'Walmart'), the system combines them into one merchant record, resolving the fragmentation caused by diverse naming conventions while maintaining accurate merchant identification.
Solution Approach 2:
The patent introduces machine learning models as an intermediary between raw transaction data and merchant identification. The models process and normalize diverse merchant names, transforming varied representations into standardized merchant identifiers. This intermediary layer enables the system to handle diverse naming conventions while maintaining high identification accuracy.
2Device complexity
If traditional methods are used to identify merchant stores in transaction data, then the process is simpler, but the accuracy of identifying and aggregating merchant data deteriorates
Solution Approach 1:
The patent performs preliminary clustering of merchant identifiers before final aggregation. By first grouping similar merchant names into clusters and then merging those clusters, the system prepares the data in advance for accurate aggregation. This preliminary organization simplifies the subsequent merging process while significantly improving identification accuracy.
Solution Approach 2:
The patent implements feedback mechanisms where the system continuously refines merchant cluster assignments based on similarity comparisons and machine learning model predictions. The feedback loop allows the system to learn from identification patterns and improve accuracy over time, maintaining a balance between process complexity and precision.
3Measurement precision
If machine-learning models are used to refine word embeddings for merchant identification, then the accuracy of merchant store identification improves, but the computational complexity and processing time increase
Solution Approach 1:
The patent segments the merchant identification process into distinct stages: initial clustering of merchant identifiers, generation of cluster names, comparison of cluster similarities, and final merging. This segmentation allows machine learning models to be applied selectively at critical decision points rather than throughout the entire process, reducing overall computational complexity while maintaining high accuracy.
Solution Approach 2:
The patent applies machine learning models partially, focusing computational resources on the most critical aspects of merchant identification where they provide the greatest value. Rather than using complex models for every aspect of processing, the system applies them strategically to refine embeddings and determine cluster similarities, achieving high accuracy with optimized computational expenditure.
Data Source
AI summary
A system including one or more processors and a computer-readable, non-transitory medium including instructions which cause at least one of the one or more processors to obtain merchant data including a plurality of merchants, obtain a set of word embeddings extracted using a large language model, refine the set of word embeddings by executing a machine-learning model using as input the merchant data to obtain a set of merchant embeddings, determine a first cluster of first merchant embeddings and a second cluster of second merchant embeddings within the set of merchant embeddings, determine a first name for the first cluster based on the first embeddings and a second name for the second cluster based on the second embeddings, and merge the first cluster and the second cluster based on a similarity of the first name and the second name to obtain a merged cluster.