Density-Based Clustering Using Spanning Trees for Real-Time Streaming Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional incremental density-based clustering algorithms face a significant slowdown when removing data, as they require expensive graph traversal operations to check for cluster separation, limiting their real-time applicability.
Innovation Solution
A new density-based clustering algorithm that uses a sliding window to process streaming data, classifies data points into cores, borders, or noise based on spatial and temporal information, and updates clusters using a spanning tree structure, eliminating the need for graph traversal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If graph traversal is used to check for cluster separation when data is removed, then cluster separation can be accurately determined, but data removal speed becomes slow
Solution Approach 1:
The patent pre-calculates and stores core expiration time points when data is inserted into the sliding window. This preliminary action allows the system to determine cluster separation by simply comparing current time with pre-stored expiration times, eliminating the need for expensive graph traversal operations during data removal.
Solution Approach 2:
The patent creates a simplified representation of cluster connectivity using spanning trees that track core expiration times. Instead of maintaining the full graph structure and performing traversal, the system uses this copied, simplified structure (spanning trees with expiration time points) to determine separation, significantly reducing computational cost while maintaining accuracy.
2Adaptability or versatility
If conventional incremental density-based clustering algorithms are used, then data clustering can be performed in streaming environment, but real-time utilization is limited due to slow data removal
Solution Approach 1:
The patent dynamically updates cluster structures by maintaining spanning trees that reflect current data relationships. When data is removed, the system dynamically adjusts cluster assignments by checking core expiration times and updating spanning trees accordingly, enabling fast real-time responses while maintaining streaming environment compatibility.
Solution Approach 2:
The patent changes the key parameter from using graph traversal operations to using pre-stored core expiration time points. This parameter change transforms the data removal process from an expensive O(n) graph traversal to a fast O(1) time comparison, enabling real-time utilization in streaming environments.
Data Source
AI summary
A density-based data clustering apparatus includes a memory storing a data clustering program, and a processor configured to execute the data clustering program, wherein the data clustering program stores input data, which is added to a sliding window for processing streaming data, together with spatial information and temporal information, removes deviation data deviating from the sliding window, classifies the input data into a core, a border, or noise according to density based on the spatial information and the temporal information, updates a cluster based on a result of classification of the input data and a result of removal of the deviation data, forms the cluster including one or more cores and borders, and updates the cluster through a spanning tree connecting cores within a critical distance by an edge.