Generalization Hierarchy Generation for Non-Numerical Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for creating generalization hierarchies struggle with data that lacks numerical meaning, often resulting in significant information loss and unnatural generalizations, particularly when using frequency-based approaches or thesauruses, which can lead to contradictory hierarchies.
Innovation Solution
A method involving a first and second generation part to generate generalization hierarchies that include shared conceptual data, using depth-first search and formatting algorithms to ensure that the generated hierarchies minimize information loss and maintain conceptual consistency, even for data without numerical meaning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If frequency-based methods are used to create generalization hierarchies, then privacy protection is improved, but information loss increases significantly for data without numerical meaning
Solution Approach 1:
The patent changes the parameter basis for hierarchy creation from frequency (numerical) to conceptual relationships (semantic). By using thesaurus-based semantic analysis and conceptual distance metrics, the system can create generalization hierarchies for non-numerical data that preserve conceptual meaning rather than relying on frequency counts, thereby reducing information loss while maintaining privacy protection.
Solution Approach 2:
The patent introduces a thesaurus as an intermediary tool to bridge the gap between data items and their conceptual relationships. The thesaurus provides semantic information and conceptual hierarchies that enable the system to understand and generalize non-numerical data meaningfully, avoiding the information loss that occurs with pure frequency-based methods.
2Stability of the object's composition
If thesaurus-based methods are used to create generalization hierarchies, then conceptual consistency is improved, but contradictory hierarchies may be generated
Solution Approach 1:
The patent implements feedback mechanisms through iterative optimization processes. The system evaluates generated hierarchies against multiple criteria including conceptual consistency, privacy protection requirements (k-anonymity), and information loss metrics. Through iterative refinement and optimization, the system adjusts hierarchy structures to eliminate contradictions while maintaining conceptual consistency, ensuring valid and reliable generalization hierarchies.
3Reliability
If generalization is applied to protect privacy, then privacy protection is improved, but data utility deteriorates due to information loss
Solution Approach 1:
The patent changes the generalization approach from frequency-based grouping to concept-based grouping using thesaurus information. This allows data to be generalized along semantically meaningful dimensions rather than arbitrary frequency-based clusters, preserving more useful information and data utility while still achieving the required privacy protection levels.
Solution Approach 2:
The patent employs dynamic optimization to find the optimal balance between privacy protection and data utility. The system can adjust generalization parameters and hierarchy depths dynamically based on the specific requirements and data characteristics, allowing flexible trade-off management between privacy and utility rather than using fixed generalization rules.
Data Source
AI summary
A technology that generates a set of generalization hierarchies that reduces information loss when generalizing any kind of data that does not necessarily have a numerical meaning. Included is a second generation part that generates a second generalization hierarchy set that satisfies a predetermined property with respect to a generalization target data set and a generalization hierarchy set (in which the generalization hierarchy set contains a generalization hierarchy including any of the generalization target data included in the generalization target data set as at least one element), and provided that M is the maximum value of the length of the generalization hierarchies included in the generalization hierarchy set, D is a predetermined integer equal to or greater than 1 and less than or equal to M, and D′ is a predetermined integer equal to or greater than D and less than or equal to M.