Additive Tree Clustering via Outlier Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional cluster validation techniques for additive trees often fail to accurately identify outliers and preserve cluster imbalances, leading to incorrect cluster divisions and invalid cluster formations due to branch division approaches.
Innovation Solution
Implementing a cluster validation process that includes identifying outliers and preserving cluster imbalances by using a user device with clustering logic to partition additive trees, evaluate partitions based on specific criteria, and iteratively refine cluster divisions until valid clusters are achieved.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If branch division approaches are used for cluster validation, then the clustering process can be simplified, but outliers cannot be accurately identified and cluster imbalances are not preserved
Solution Approach 1:
The patent extracts outliers from the clustering process by identifying and removing them before performing branch division. This is achieved through computing outlier scores based on distance metrics and sequentially removing points with highest outlier scores, thereby preventing outliers from distorting cluster formation while maintaining simplicity of the overall process
Solution Approach 2:
The patent performs preliminary actions by pre-identifying and removing outliers before the main cluster validation process. This preliminary outlier removal ensures that subsequent branch division operations work only with genuine cluster members, preserving cluster imbalances and improving validation accuracy without adding complexity to the core algorithm
2Device complexity
If branch division approaches are used for cluster validation, then computational complexity is reduced, but cluster imbalances are lost leading to incorrect cluster divisions
Solution Approach 1:
The patent extracts and removes outlier points before cluster validation to prevent them from causing incorrect cluster divisions. By extracting these problematic points using distance-based outlier scoring and sequential removal, the remaining data forms valid clusters that preserve natural imbalances without requiring complex validation algorithms
Solution Approach 2:
The patent performs preliminary outlier detection and removal to ensure subsequent clustering operations produce reliable results. This preliminary action of identifying and removing outliers based on distance metrics establishes a clean dataset for branch division, ensuring cluster validity without adding computational complexity to the main validation process
3Productivity
If conventional cluster validation is applied to additive trees, then the process can be completed quickly, but decision errors result in incorrect number of clusters
Solution Approach 1:
The patent extracts outliers using distance-based scoring and sequential removal before applying branch division. This extraction of problematic outlier points prevents decision errors in determining the number of clusters, allowing conventional validation methods to proceed quickly on cleaned data with accurate cluster count identification
Solution Approach 2:
The patent performs preliminary outlier removal to eliminate sources of decision error before cluster validation. By pre-identifying and removing outliers based on distance metrics, the subsequent validation process operates on clean data, maintaining high processing speed while accurately determining the correct number of clusters
Data Source
AI summary
A computing device-implemented method includes receiving an additive tree; assigning data associated with the additive tree to one or more initial clusters; partitioning the additive tree into one or more pairs of additive sub-trees corresponding to one or more binary segmentations; computing a set that includes partitions resulting from a combination of the one or more initial clusters and the one or more pairs of additive sub-trees; evaluating one or more partitions of the set with one or more cluster validation criteria; storing one or more evaluation results for the one or more partitions; selecting at least one partition from the one or more partitions of the set that satisfies the one or more cluster validation criteria, where the at least one partition is associated with an optimal evaluation result; and removing at least one of the binary segmentations that corresponds to the at least one partition.