Interactive Data Exploration Using Similarity-Based Visual Layouts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data management and analysis systems struggle to efficiently manage and analyze large datasets with complex relationships, particularly in multi-dimensional environments, leading to inefficiencies in data visualization and retrieval.
Innovation Solution
The system employs an interactive data exploration method that utilizes graphical objects to represent records, adjusts their positions based on similarity scores, and applies Metric Multi-Dimensional Scaling techniques to optimize their arrangement, enabling dynamic data subset generation and real-time analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional data management systems are used to handle large datasets with complex relationships, then data storage capacity is sufficient, but data visualization efficiency and retrieval speed deteriorate
Solution Approach 1:
The patent replaces traditional mechanical data retrieval methods (SQL queries, manual filtering) with a force-directed graph system that uses simulated physical forces (attraction and repulsion) to automatically organize and visualize data relationships. This substitution enables efficient exploration of large datasets through intuitive visual interactions rather than complex query formulation.
Solution Approach 2:
The patent introduces graphical objects as intermediary elements between the user and the underlying data records. These graphical objects serve as visual mediators that represent data entities and their relationships, allowing users to interact with complex data through simplified visual metaphors rather than directly manipulating raw data or complex query structures.
2Loss of information
If more fields are selected for analysis, then data exploration depth increases, but computation time for similarity scores increases
Solution Approach 1:
The patent implements incremental computation where similarity scores are calculated and updated progressively as users interact with the visualization, rather than computing all possible comparisons upfront. The system computes similarity metrics on-demand based on user selections and current view state, performing only the necessary calculations required for the current exploration step.
Solution Approach 2:
The patent pre-computes and stores similarity metrics between data records before user interaction begins. These pre-calculated similarity scores are then reused and adjusted during interactive exploration, avoiding redundant computations when users filter or select different field combinations.
3Quantity of substance
If graphical objects are densely arranged to show more data, then data density increases, but visual clarity and user interaction difficulty increase
Solution Approach 1:
The patent implements a dynamic layout system where graphical objects automatically adjust their positions based on simulated physical forces. Objects with similar characteristics are attracted together while dissimilar objects are repelled, creating organic clusters that maintain visual clarity even when displaying large numbers of records. The layout continuously adapts to user interactions and filtering operations.
Solution Approach 2:
The patent uses force-directed positioning in a two-dimensional visual space to represent multi-dimensional data relationships. By mapping complex data similarities and differences onto spatial positions, the system enables users to perceive data structure and relationships through spatial arrangement rather than traditional tabular formats.
Data Source
Figure 1A
Figure 1B
Figure 1C
AI summary
The present invention relates to a method in a data record selection interface for extracting data records having similar characteristics. The method comprises displaying, in the selection interface, records of a data set as graphical objects, extracting a plurality of fields from the displayed records and displaying each extracted field in the interface as selectable entities, receiving selections of a plurality of fields, calculating vectors of similarities for at least the selected plurality of fields, each vector of similarities includes a similarity score for each pair of values of the records of the data set, determine a combined similarity score for the selected plurality of fields and for each pair of records, each combined similarity score being a value representing the similarities between the selected fields of the data record pair, calculating positions for each graphical object by matching distances between each pair of graphical objects, each distance between a pair of graphical objects being based on the combined similarity score for the corresponding record pair, displaying the graphical objects according to the calculated positions, receiving selection of a plurality of graphical objects, the selection being of an area in the selection interface including the plurality of graphical objects, and extracting and assembling the records underlying the selection into a data set.