Cluster Interestingness Ranking for Data Mining

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Manual analysis of large numbers of clusters produced by data mining can be time-consuming and may lead to missing interesting or unexpected patterns, especially in large datasets like news articles or litigation documents, where expected clusters do not provide additional insights.

Innovation Solution

Computing interestingness measures for clusters using techniques like K-means clustering, identifying connected concepts, and calculating facet connection information and entropy to rank clusters based on their interestingness, allowing users to focus on unexpected associations.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual analysis of all clusters is performed, then complete coverage of all clusters is achieved, but time consumption increases significantly

Engineering Contradiction:

Improveanalysis completenessVSAvoidtime consumption

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs self-evaluation by automatically computing interestingness measures for each cluster using algorithms that assess concept connectivity and entropy, enabling the system to prioritize clusters without human intervention

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent transforms the analysis task by introducing new parameters (interestingness measures including concept connectivity and entropy) that automatically differentiate clusters, allowing the system to identify promising clusters through quantitative criteria rather than manual review

Inventive Principle:

Principle #35Parameter changes

2Reliability

If all clusters are reviewed manually, then no interesting patterns are missed, but the process becomes inefficient and scalable only to small datasets

Engineering Contradiction:

Improvepattern detection reliabilityVSAvoidanalysis efficiency

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The patent introduces an intermediary computational layer that calculates interestingness measures as a bridge between raw clustering results and human analysis, automatically filtering and ranking clusters to guide human attention toward the most promising candidates

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system replaces the mechanical process of manual cluster review with automated computational algorithms that calculate concept connectivity and entropy, substituting human labor with machine-based interestingness assessment

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Quantity of substance

If clustering is performed on large datasets, then more comprehensive data coverage is achieved, but the number of clusters increases making analysis overwhelming

Engineering Contradiction:

Improvedata coverageVSAvoidcluster management complexity

Core Design Contradiction:

Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent segments the large set of clusters by computing interestingness measures for each one, dividing the overwhelming whole into prioritized subsets that can be managed and analyzed in order of their interestingness scores

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS9230015B2Deriving an interestingness measure for a cluster

Publication Date: 2016.01.05 MICRO FOCUS LLC

AI summary

Groups of connected concepts are identified for a cluster. At least one interestingness measure is computed for each of the groups of connected concepts in the cluster. An interestingness measure for the cluster is derived based on the interestingness measures for the corresponding groups of connected concepts.