Audience Recommendation via Graph Embeddings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for targeted advertisements face challenges in effectively selecting audiences, particularly as the scale increases, as they often rely on manual processes that are inefficient and cannot be solely supported by market research, and struggle to uncover non-obvious relationships between merchants and consumers.
Innovation Solution
The method employs node similarity in combined contextual graph embeddings, leveraging transaction data and auxiliary data to characterize merchants and audiences, forming cohorts and revealing opportunities for growth by generating merchant and audience embeddings and analyzing them within a heterogeneous information network to identify relationships and provide audience recommendations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If manual processes are used for audience selection in targeted advertisements, then the process is simple to implement, but the efficiency and scalability deteriorate as the scale increases
Solution Approach 1:
The patent replaces manual mechanical processes with an automated computational system that uses graph embeddings and machine learning algorithms to analyze transaction data and identify audience-merchant relationships, thereby eliminating the inefficiency of manual scaling while maintaining ease of use through automated processing
Solution Approach 2:
The system enables self-service by automatically processing transaction data, generating graph embeddings, and identifying target audiences without requiring manual intervention, allowing the system to scale efficiently while remaining easy to operate through automated decision-making
2Ease of manufacture
If traditional market research methods are used, then the approach is straightforward, but the ability to uncover non-obvious relationships between merchants and consumers deteriorates
Solution Approach 1:
The patent transforms traditional flat market research data into multi-dimensional graph embeddings that capture complex relationships between merchants, consumers, and transactions, enabling the discovery of non-obvious patterns and connections that traditional methods cannot detect while maintaining analytical straightforwardness
3Measurement precision
If comprehensive transaction data analysis is performed to identify non-obvious relationships, then the accuracy of audience selection improves, but the computational complexity and data processing requirements worsen
Solution Approach 1:
The patent segments the comprehensive transaction data into structured graph embeddings that represent merchants, consumers, and their relationships as discrete nodes and edges, enabling efficient computational processing while preserving the accuracy needed to identify precise audience-merchant matches
Data Source
AI summary
A method for audience recommendation using node similarity in combined contextual graph embeddings can include receiving a merchant identifier of a merchant and generating one or more merchant tags describing merchant data corresponding to the merchant. A set of audience embeddings can be generated from a set of audience auxiliary data using an audience taxonomy and a set of merchant embeddings can be generated from the merchant data relating to the merchant using the one or more merchant tags. The set of audience embeddings and the set of merchant embeddings are used to produce a heterogenous information network of combined audience data and merchant data, which is then analyzed to identify relationships between each audience and the merchant. A score for one or more audiences can be determined based on the relationships between the one or more audiences and the merchant.