Interactive Graph Mining System for Network Visualization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Mining large complex networks of information nodes is challenging due to the complexity and cost of visual interaction, as well as the difficulty for end-users to understand community detection algorithm results, especially when dealing with datasets involving more than a few dozen nodes.
Innovation Solution
A user-guided data mining system that includes a database for network graphs, node, and edge attributes, and a data mining module to request user feedback for attribute selection, visualization, and ranking of graph mining algorithms based on user input, allowing users to interactively choose algorithms and visualize network graphs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If community detection algorithms are applied to large networks, then the community structure can be found, but the results are difficult for end-users to understand and the computational cost is high
Solution Approach 1:
The patent introduces an interactive visualization system as an intermediary between the community detection algorithm and the end-user. The system displays network graphs with detected communities visually represented, allowing users to see the results in an intuitive graphical format rather than raw data. Users can interact with the visualization to explore community structures, making the algorithm results accessible and understandable without requiring users to interpret complex computational outputs directly
Solution Approach 2:
The system implements feedback mechanisms where users can provide input about the visualized results, such as selecting specific communities for further analysis or adjusting visualization parameters. This feedback loop allows users to guide the analysis process and refine results according to their needs, making the complex algorithm outputs adaptable to user-specific understanding requirements
2Adaptability or versatility
If users want to change parameters or algorithms in community detection, then different analysis results can be obtained, but the user needs to re-run the entire algorithm all over again
Solution Approach 1:
The system performs preliminary actions by pre-computing and storing intermediate results from the community detection algorithm, such as community assignments, modularity values, and network properties. When users want to change parameters or algorithms, the system can leverage these pre-computed results to quickly generate new analyses without re-running the entire computationally expensive algorithm from scratch, significantly reducing the time penalty for parameter adjustments
Solution Approach 2:
The system implements dynamic behavior where the analysis can be incrementally updated as users change parameters or select different algorithms. Rather than static re-execution, the system dynamically adjusts the analysis based on user input, maintaining state information that allows for efficient updates and comparisons between different algorithmic approaches without complete re-computation
3Productivity
If visual interaction with the network is used to aid mining, then the mining process can be improved, but the process becomes complicated and expensive
Solution Approach 1:
The patent segments the visual mining system into distinct functional modules: network data processing, community detection algorithm execution, result visualization, and user interaction handling. Each module performs a specific function and can be independently optimized or configured. This segmentation reduces overall system complexity by breaking down the complicated visual mining process into manageable, well-defined components that work together systematically
Data Source
AI summary
A user guided data mining system includes: a database storing a network graph comprising a plurality nodes and edges, node attributes for the nodes, and edge attributes for the edges; and a data mining module configured to: request feedback from a user indicating at least one of the node attributes, visualize the network graph considering the indicated attributes, request feedback from the user indicating a part of the visualized network graph and a class of graph mining algorithm, and rank a plurality of different graph mining algorithms of the indicated class based on the indicated part.