Content Recommendation System Using Behavioral Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional customized service systems face challenges in providing personalized recommendations on devices like TVs, as they struggle to individually identify users and classify their viewing histories, leading to impersonalized and inappropriate content suggestions due to mixed user data.
Innovation Solution
A content-providing method and system that clusters user behavioral data into representative types and maps these types to specific time intervals, allowing for customized service delivery without user identification, using a use history analyzer and time interval identifier to store and manage data effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If user identification information is used for personalized recommendations, then recommendation accuracy is improved, but system complexity and user privacy requirements increase
Solution Approach 1:
The patent extracts and removes user identification information from the recommendation system. Instead of using user IDs to track viewing histories, the system processes anonymous behavioral data, extracting only the necessary viewing pattern information while discarding personal identifiers. This resolves the contradiction by maintaining recommendation accuracy through pattern analysis without requiring complex user identification mechanisms.
Solution Approach 2:
The patent creates representative type clusters that serve as copies or models of user behavior patterns. Instead of storing and processing individual user identification data, the system generates anonymized representative profiles that capture essential viewing preferences. These representative types enable personalized recommendations without needing actual user identification information, thus reducing system complexity while maintaining accuracy.
2Quantity of substance
If viewing histories of multiple users are stored in one device, then data storage efficiency is improved, but recommendation personalization deteriorates due to mixed user data
Solution Approach 1:
The patent segments mixed user viewing histories into distinct representative type clusters. By clustering similar viewing behaviors together and identifying representative types for each cluster, the system effectively divides the mixed data into meaningful segments. This allows efficient storage of aggregated data while maintaining the ability to provide personalized recommendations, as each representative type captures the essence of a specific user segment's preferences.
Solution Approach 2:
The patent transforms raw viewing history data into representative type parameters through clustering analysis. By changing the data representation from individual user records to aggregated type parameters, the system achieves both storage efficiency and personalization capability. The representative types serve as compressed parameter sets that encode multiple users' preferences without storing all original data, resolving the contradiction between storage efficiency and recommendation precision.
3Adaptability or versatility
If user identification is implemented in family devices like TVs, then personalized service is improved, but ease of operation and user experience deteriorate due to identification difficulties
Solution Approach 1:
The patent implements a self-service approach where the system automatically performs clustering analysis and generates representative type profiles without requiring user intervention for identification. The system autonomously processes anonymous viewing data, creates behavioral clusters, and delivers personalized recommendations without asking users to log in or provide identification. This maintains personalized service capability while preserving ease of operation, as users simply view content without any identification steps.
Data Source
AI summary
A content-providing method and system, including identifying a representative type cluster by clustering content related to behavioral data which represents a use history of a user, according to type of the content, mapping the representative type cluster to a time interval, and storing the representative type cluster and the time interval.