Item-Level Interaction Categorization Using Machine Learning Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current interaction categorization methods, primarily based on merchant category codes (MCC), are insufficient for accurately representing the diverse range of items sold by a single merchant, and they fail to effectively categorize interactions at the item level, especially in online or card-not-present transactions.
Innovation Solution
The use of machine-learning-based techniques to classify and categorize items by training a machine-learning model to learn associations between item and/or merchant data and corresponding item categories, enabling accurate categorization at the item level.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If merchant category codes (MCC) are used for categorization, then categorization can be performed at the merchant level, but it fails to accurately represent the diverse range of items sold by a single merchant
Solution Approach 1:
The patent segments the categorization process from the merchant level to the item level. Instead of assigning a single category to a merchant, the system creates separate category assignments for each item sold by the merchant. This segmentation enables precise tracking of diverse items while maintaining the merchant-level framework, directly resolving the contradiction between categorization accuracy and system complexity.
Solution Approach 2:
The patent adds a new dimension to the categorization system by introducing item-level categorization as a second layer beyond merchant-level categorization. This dimensional expansion allows the system to capture both the broad merchant category and the specific item category simultaneously, achieving high precision without overwhelming complexity through structured multi-dimensional classification.
2Loss of information
If broad merchant categories are used, then categorization can be performed simply, but it is insufficient in capturing the diverse range of items sold by a single merchant
Solution Approach 1:
The system segments item-level information from merchant-level information, creating distinct category assignments for each item while preserving the merchant category structure. This segmentation prevents loss of item diversity information by explicitly capturing it at the item level, while maintaining operational simplicity through the hierarchical organization of categories.
Solution Approach 2:
The patent applies local quality by allowing different levels of categorization detail for different entities. Merchant categories provide broad classification for operational simplicity, while item categories provide detailed classification for capturing item diversity. Each level operates with appropriate granularity, resolving the contradiction between information completeness and operational ease.
3Adaptability or versatility
If traditional categorization methods are used, then the system can operate with existing infrastructure, but they are not applicable to online or card-not-present interactions
Solution Approach 1:
The patent creates a universal categorization framework that handles multiple interaction types (in-person, online, card-not-present) through the same item-level categorization mechanism. This multi-functional approach enables the system to adapt to various transaction contexts without requiring separate systems, achieving high versatility while managing complexity through a unified categorization structure.
Data Source
AI summary
A method for categorization may initiate a local instance of a widget on a device of a user, the widget including a first model. The method for categorization may include retrieving, by the widget, a first data set from the device of the user, the first data set associated with one or more datum of an active user session. The method for categorization may include scoring, by the model, a first item based at least in part on the first data set. The method for categorization may include labeling the first item based on a result of the scoring. The method for categorization may include sending, by the widget, a result of the labeling to a remote server.