Heterogeneous Data Clustering via Grid-Based Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional clustering methods face challenges with increasing data sizes, requiring significant calculations, losing characteristic information when reducing clusters, and struggling with scattered and multi-dimensional data, especially in systems like building energy management systems that use diverse sensors.
Innovation Solution
A heterogeneous data cluster generation apparatus and method that receives time series data from different sensors, merges them into ordered pairs, and uses a grid-based coordinate system with indices to efficiently cluster data, reducing calculation complexity and effectively handling massive and scattered data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the number of clusters is increased to accurately identify data characteristics, then clustering accuracy is improved, but the amount of calculation required increases significantly
Solution Approach 1:
The patent segments the calculation process by dividing data into multiple clusters in parallel, with each processing unit handling a specific cluster. This allows accurate identification of data characteristics across multiple clusters while distributing the computational load, thereby maintaining clustering accuracy without proportionally increasing overall calculation time.
2Productivity
If the number of clusters is reduced to decrease calculation amount, then calculation efficiency is improved, but data lose their original characteristic information
Solution Approach 1:
The patent segments the data space into multiple clusters, each processed by dedicated processing units. This segmentation enables the system to maintain more clusters than traditional sequential methods would allow, preserving data characteristic information while achieving parallel processing efficiency.
Solution Approach 2:
The patent performs preliminary actions by pre-defining cluster boundaries and characteristics before actual data processing. This allows the system to efficiently assign data points to appropriate clusters without performing exhaustive calculations, thereby maintaining data characteristics while improving calculation efficiency.
3Productivity
If hierarchical algorithm such as K-D Tree is used to avoid distance calculation for all clusters, then calculation amount is reduced, but the number of nodes to be searched increases geometrically when dimensions exceed 10
Solution Approach 1:
The patent segments the high-dimensional space into multiple manageable clusters, each handled by dedicated processing units. This segmentation avoids the geometric increase in node searches by distributing the search space across multiple processors, each dealing with a subset of dimensions and data points.
Solution Approach 2:
The patent introduces a parallel processing dimension by distributing cluster processing across multiple processing units. This transforms the single-threaded hierarchical search into a multi-threaded approach, effectively adding a computational dimension that prevents the geometric increase in search complexity.
4Ease of operation
If conventional clustering method is used for scattered data, then clustering can be performed, but different clustering results are produced every time requiring frequent re-clustering
Solution Approach 1:
The patent performs preliminary actions by pre-defining cluster characteristics and boundaries based on the scattered data distribution. This preliminary setup ensures that subsequent clustering operations produce consistent results, eliminating the need for frequent re-clustering while maintaining the ability to handle scattered data effectively.
Data Source
AI summary
The present invention relates to a heterogeneous data cluster generation apparatus and method and a data clustering method and apparatus, and more particularly, to a data clustering method and apparatus which cluster data measured by different sensors into a number of groups. Aspects of the present invention provide an apparatus and method for generating clusters by putting together heterogeneous data which are values measured by different types of sensors. Aspects of the present invention also provide an apparatus and method for generating clusters by setting indices in order to effectively cluster multi-dimensional data, massive data, or scattered data.