Database Model Clustering via Edge Betweenness and Hub Nodes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Large and complex enterprise data models and database schemas are difficult to understand due to their complexity, especially when undocumented, making tasks like analytical reporting and schema matching labor-intensive and costly, requiring domain experts or consultants.
Innovation Solution
A method for clustering database objects in a database model using edge betweenness values and degree calculations to identify linking and hub nodes, allowing for the formation of clusters and generation of meaningful cluster names, which simplifies the understanding and analysis of data models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If domain experts or consultants manually analyze large enterprise data models and schemas, then understanding and comprehension of the data model is achieved, but the process becomes very labor-intensive and costly
Solution Approach 1:
The system performs self-service by automatically analyzing the data model itself through graph-based algorithms. The clustering algorithm processes the schema representation autonomously, identifying topic areas and generating summaries without requiring external domain experts, thus eliminating the labor-intensive manual analysis process while preserving comprehensive understanding of the data model
Solution Approach 2:
The patent replaces the mechanical manual analysis process with an automated computational system. Graph algorithms calculate edge betweenness values and identify hub nodes programmatically, substituting human expert effort with algorithmic processing that achieves the same analytical goals more efficiently and at lower cost
2Ease of manufacture
If the data model is undocumented or poorly documented, then the data model can be acquired more easily, but understanding and analyzing the model becomes much more difficult
Solution Approach 1:
The system applies local quality by generating targeted summaries for specific clusters rather than requiring comprehensive documentation of the entire data model. Each cluster receives a localized summary based on its hub node, providing just enough contextual information to understand that specific portion of the schema without needing full documentation of all components
Solution Approach 2:
The patent segments the complex data model into smaller, manageable clusters based on edge betweenness analysis. By dividing the schema into topic-area-specific groups, the system makes understanding easier while maintaining ease of acquisition, as each segment can be analyzed independently rather than requiring comprehension of the entire monolithic schema
3Device complexity
If clustering is applied to reduce data model complexity, then understanding and analysis becomes easier, but the computational processing required to determine edge betweenness values and identify hub nodes increases
Solution Approach 1:
The algorithm segments the computational task by processing the data model in iterations, identifying one hub node per cluster at a time. This divides the overall computational burden into manageable steps, where each iteration focuses on finding hub nodes for remaining unclustered nodes, making the processing more efficient than attempting to analyze the entire schema simultaneously
Solution Approach 2:
The system extracts and removes identified hub nodes and their connected components from the graph after each iteration. By taking out the clustered portions from further processing, the algorithm reduces the computational scope for subsequent iterations, preventing redundant calculations and optimizing the overall processing power required
Data Source
AI summary
A method for data model clustering is provided herein. A first representation of a data model may be received. Edge betweenness values may be determined for respective nodes in the first representation. At least one node in the first representation may be identified as a linking node based on the respective edge betweenness values. One or more linking nodes may be removed from the first representation, thereby forming at least a first cluster and a second cluster. Degrees for the respective remaining nodes may be calculated. Respective hub nodes may be identified for the respective clusters based on the respective degrees in the clusters. Respective descriptions may be generated for the respective clusters based on the respective hub nodes. A clustered representation of the first representation may be stored with the clusters and their respective descriptions.