User Embedding Expansion for Accurate Look-Alike Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional analytics engines struggle to efficiently analyze user profile and attribute data to represent users in a uniform manner, leading to inaccurate and imprecise user segment expansions, which wastes computing resources and bandwidth by providing content to uninterested users.
Innovation Solution
A user embeddings system utilizing a trained neural network to generate uniform user embeddings, enabling the automatic expansion of user segments by identifying look-alike or holistically-similar users based on user behavior and traits, and providing a graphical user interface for visualization and customization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional analytics engines use static rule-based segments to characterize users, then the system can operate with simple processing logic, but the user segment accuracy deteriorates and computing resources are wasted on uninterested users
Solution Approach 1:
The patent transforms user data from static attributes to dynamic embeddings by changing the representation parameters. Neural networks encode user profile data, behavioral data, and attribute data into continuous vector embeddings, enabling nuanced similarity calculations that accurately identify look-alike users while optimizing resource allocation.
Solution Approach 2:
The patent replaces the mechanical rule-based segmentation system with a neural network-based embedding system. Instead of using rigid if-then rules to categorize users, the system uses learned representations that capture complex user characteristics and behaviors, improving segmentation accuracy without proportional increases in computational waste.
2Quantity of substance
If conventional analytics engines expand user segments using inefficient methods, then the system can increase segment size, but the manufacturing precision of user segment expansion deteriorates
Solution Approach 1:
The patent changes the expansion methodology from rule-based addition to embedding-based similarity search. By representing users as vectors in continuous space, the system can efficiently identify and add users with similar characteristics to existing segments, increasing segment size while maintaining high precision through mathematical similarity metrics.
3Productivity
If conventional analytics engines analyze user profile data without uniform representations, then the system can process diverse data types, but the device complexity increases and analysis efficiency decreases
Solution Approach 1:
The patent creates a universal embedding representation that can uniformly encode diverse user data types including profile information, behavioral patterns, and attributes. This single representation framework handles multiple data types consistently, improving analysis efficiency without proportionally increasing system complexity through standardized processing pipelines.
Data Source
AI summary
The present disclosure relates to systems, non-transitory computer-readable media, and methods for expanding user segments automatically utilizing user embedding representations generated by a trained neural network. For example, a user embeddings system expands a segment of users by identifying holistically similar users from uniform user embeddings that encode behavior and/or realized traits of the users. Further, the user embeddings system facilitates the expansion of user segments in a particular direction and focus to improve the accuracy of user segments.