Ontology-Based Recommendation System for Sparse User Data Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional recommendation systems face inaccuracies due to user data sparsity, leading to incorrect clustering and recommendation results, as most users purchase or interact with only a few products, resulting in low data density.
Innovation Solution
A recommendation system that utilizes an ontology database with a hierarchy structure to convert user data from lower hierarchy levels to higher levels, reducing sparsity and improving clustering accuracy by generating user data that records entities on different hierarchy levels, allowing for more accurate clustering and recommendation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If user data is collected and clustered directly without transformation, then the recommendation system can operate with simple processing, but the sparsity of user data causes incorrect clustering and reduces recommendation accuracy
Solution Approach 1:
The patent transforms the flat user data structure into a hierarchical ontology structure with multiple levels. By mapping user interactions to entities at different hierarchy levels (e.g., product → product category → department), the system adds a dimensional layer that aggregates sparse data points into denser hierarchical representations, enabling more accurate clustering despite original data sparsity
Solution Approach 2:
The patent changes the parameter representation by converting user interaction data from direct product-level parameters to hierarchical ontology parameters. This transformation modifies the data structure from sparse individual item interactions to denser hierarchical category representations, fundamentally changing how similarity is measured and clustered
2Measurement precision
If user data is transformed to higher hierarchy levels to reduce sparsity, then clustering accuracy improves, but the system complexity increases due to ontology database requirements
Solution Approach 1:
The patent performs preliminary action by pre-building the ontology database with hierarchical entity relationships before user data processing. This pre-computed hierarchical structure is then reused during clustering operations, avoiding the need to dynamically compute hierarchical relationships during each recommendation task, thus managing complexity through upfront preparation
Solution Approach 2:
The ontology database serves as an intermediary layer between raw user interaction data and the clustering algorithm. Instead of directly processing sparse user data or implementing complex hierarchical computations in the clustering algorithm itself, the system uses the pre-structured ontology as a mediator that transforms and aggregates data into suitable representations for clustering
Data Source
AI summary
A recommendation method includes providing an ontology database, in which the ontology database includes a plurality of entities, and the entities are arranged in an ontology hierarchy structure with N hierarchy levels; storing a plurality of jth level user data respectively corresponding to a plurality of users; generating a plurality of kth level user data according to the jth level user data respectively; clustering the kth level user data; and recommending the entities in the ontology database to the users according to a clustering result.