User Interest Determination via Classification Behavior Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current recommendation systems for Internet products face challenges in accurately determining user interest for information items due to limitations in reflecting inherent user preferences and dealing with noise, especially in scenarios with many users, where the cost of calculating similarity matrices is high and data sparseness is a problem.
Innovation Solution
A method and apparatus that utilize classification behavior information representations and vectorized information to determine user interest by obtaining and processing behavior classification data, including click, browse, purchase, and comment actions, to create a more accurate representation of user preferences, incorporating historical behavior data and relationship data to form comprehensive interest determinations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If similarity matrix calculation is used to determine user interest, then recommendation accuracy can be improved, but computational cost increases significantly
Solution Approach 1:
The patent segments user behaviors into multiple behavior classifications (e.g., click, browse, purchase, comment) and processes each classification separately to create behavior-specific vector representations. This segmentation allows the system to avoid calculating full similarity matrices for all user-item pairs while maintaining accurate interest determination by focusing computations on relevant behavior types.
Solution Approach 2:
The patent transforms user behavior data into vector representations with specific dimensions corresponding to different behavior classifications. By changing the parameter representation from raw behavior data to structured vectors, the system enables efficient similarity computation without requiring exhaustive matrix calculations, thus reducing computational cost while preserving measurement precision.
2Measurement precision
If more user behavior data is collected to improve recommendation accuracy, then user interest determination improves, but data sparseness problems worsen
Solution Approach 1:
The patent merges multiple behavior classifications (click, browse, purchase, comment) into a unified vector representation framework. By combining different types of user behaviors into a comprehensive behavior vector, the system enriches the data representation and reduces sparseness issues that would arise from analyzing single behavior types in isolation.
Solution Approach 2:
The patent introduces behavioral dimensionality by creating vector representations where each dimension corresponds to a specific behavior classification. This dimensional transformation converts sparse scalar behavior data into dense vector representations, enabling the system to capture nuanced user preferences while mitigating data sparseness through the added structural dimensions.
3Measurement precision
If multiple behavior classifications are processed separately, then user preference representation accuracy improves, but system complexity increases
Solution Approach 1:
The patent creates a universal vector representation framework that handles multiple behavior classifications through a unified processing mechanism. The same vectorization and similarity computation methods are applied across different behavior types, allowing the system to maintain high representation accuracy while avoiding the complexity of separate processing pipelines for each behavior classification.
Data Source
AI summary
The present disclosure provides a method and an apparatus for determining interest of a user for an information item, a computer device, and a computer-readable storage medium. The method includes obtaining, according to behaviors of a plurality of behavior classifications of a target user, a classification behavior information representation including vectorized information of behaviors of each behavior classification of the target user; obtaining a vectorized information representation of a candidate information item; and determining interest of the target user for the candidate information item according to the classification behavior information representation of the behaviors of the target user and the vectorized information representation of the candidate information item.