Tiered Hierarchical Clustering for Binary and One-Hot Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional distance-based clustering methods struggle with binary and one-hot encoded data, often forming clusters with dissimilar data points and generating significant amounts of orphaned data, lacking flexibility in determining clustering tiers and thresholds.
Innovation Solution
A hierarchical clustering approach using different feature subsets at various tiers, employing a machine learning model to cluster data points into multiple clusters based on similarity values computed using distinct feature subsets, thereby reducing orphaned data and improving cluster accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If conventional distance-based clustering methods are used, then the clustering process is simple, but the clustering accuracy deteriorates and orphaned data increases
Solution Approach 1:
The patent segments the feature set into multiple subsets (first subset, second subset, etc.) and applies different subsets at different clustering tiers. This segmentation allows the system to capture different aspects of data similarity at different levels, improving clustering accuracy for binary and one-hot encoded data while maintaining a structured processing approach.
Solution Approach 2:
The patent introduces a hierarchical tiered structure with multiple clustering levels (first tier, second tier, etc.), adding a dimensional aspect to the clustering process. Each tier uses different feature subsets and distance thresholds, transforming the single-level clustering into a multi-dimensional approach that significantly reduces orphaned data and improves accuracy.
2Stability of the object's composition
If conventional hierarchical clustering techniques are used, then the clustering structure is organized, but the flexibility in determining clustering tiers and features deteriorates
Solution Approach 1:
The patent implements a dynamic hierarchical clustering system where different feature subsets are selectively applied at different tiers based on the data characteristics. The system can adaptively choose which feature subsets to use at each tier and adjust distance thresholds dynamically, providing both structured organization and flexible adaptability to various data types and clustering requirements.
Solution Approach 2:
The patent applies different feature subsets to different clustering tiers, making each tier specialized for specific aspects of data similarity. The first tier might use one subset of features while the second tier uses another subset, allowing each part of the hierarchical structure to have optimized local quality for its specific clustering objective.
3Productivity
If single feature set clustering is used, then the processing is efficient, but the ability to capture diverse data relationships deteriorates
Solution Approach 1:
The patent segments the feature set into multiple subsets and applies them at different hierarchical tiers. This segmentation enables the system to capture diverse data relationships by examining different feature combinations at each tier, improving reliability while maintaining processing efficiency through the structured hierarchical approach.
Solution Approach 2:
The patent applies feature subsets selectively at different tiers rather than using all features uniformly. Each tier uses a partial set of features appropriate for its specific clustering objective, avoiding the computational burden of processing all features at every tier while still capturing comprehensive data relationships across the hierarchy.
Data Source
AI summary
Techniques for hierarchical clustering with tiered specificity are disclosed herein. An example computer-implemented method includes receiving data points that each include data corresponding to a feature set. The example computer-implemented method further includes applying, a machine learning model to the data points to: cluster (i) a first portion of the data points into a first cluster set based on similarity values computed using a first subset of the feature set and (ii) a second portion of the plurality of data points into a second cluster set based on similarity values computed using a second subset of the feature set that is different from the first subset. The example computer-implemented method further includes generating a data object indicating a course of action for an entity associated with a first data point based on the first data point being included in the first cluster set or the second cluster set.