Clustering Coefficient Adaptive Clustering Algorithm
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing clustering algorithms only consider similarities between data points and a core element or adjacent elements, failing to account for similarities between elements within a category, leading to inaccurate data clustering.
Innovation Solution
A clustering coefficient-based adaptive clustering method that uses a preliminary clustering step to establish association relations between data points and categories, followed by a self-learning clustering step for segmentation and merging, calculating intra-category and inter-category similarities to refine cluster assignments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If existing clustering algorithms only consider similarities between data points and core elements or adjacent elements, then the algorithm complexity is reduced and computation is simplified, but the clustering accuracy deteriorates because pairwise similarities between elements in the category are not considered
Solution Approach 1:
The algorithm enables data points to self-evaluate their pairwise similarities with other elements in the same category through the clustering coefficient calculation. Each data point automatically determines its local clustering coefficient by examining its neighborhood relationships, allowing the system to self-assess clustering quality without external intervention or simplified assumptions
Solution Approach 2:
The clustering coefficient serves as a feedback mechanism that evaluates the quality of clustering by measuring pairwise similarities between elements. The algorithm uses this feedback to iteratively refine cluster assignments, comparing the clustering coefficient against a threshold to determine whether data points should be merged or separated, thereby continuously improving clustering accuracy
2Measurement precision
If pairwise similarities between elements in the category are considered through clustering coefficient calculation, then clustering accuracy is improved, but the computational complexity and time consumption increase
Solution Approach 1:
The algorithm segments the computational task into two phases: a preliminary clustering phase that creates initial categories using simpler methods, and a refinement phase that applies clustering coefficient calculations only within each category. This segmentation reduces overall computation time by avoiding exhaustive pairwise comparisons across the entire dataset while still achieving accurate clustering through localized coefficient calculations
Solution Approach 2:
The algorithm performs preliminary clustering before applying the computationally intensive clustering coefficient calculations. By first organizing data into rough categories using faster methods, the system reduces the search space for subsequent detailed analysis, allowing pairwise similarity measurements to be performed only on smaller, pre-grouped subsets of data rather than the entire dataset
3Productivity
If data points are clustered based only on similarity to core elements, then the clustering process is simpler and faster, but the ability to identify true category structures deteriorates because relationships between all elements are not captured
Solution Approach 1:
The algorithm merges multiple similarity assessment approaches: it combines the efficiency of core-element-based clustering with the accuracy of pairwise similarity measurement through clustering coefficients. By integrating both methods, the system maintains fast initial clustering while enhancing category identification accuracy through the additional insight provided by element-to-element similarity measurements
Data Source
AI summary
A clustering coefficient-based adaptive clustering method, according to the categories of extracted data point pairs and the magnitude of association relations between data points extracted each time, determining to which category the data points belong, and the number of the categories, and establishing association relations between the data points and association relations between the categories; pre-segmenting each category and calculating the intra-category similarities of two sub-categories and inter-category similarities of the two sub-categories, judging whether the two pre-segmented sub-categories satisfy a segmentation condition, if so, then accepting the pre-segmentation; if not, then canceling the pre-segmentation; calculating the intra-category similarities and inter-category similarities of two categories having an association relation, judging whether the two categories satisfy a merging condition, if so, then merging the two categories to generate a new category; if not, then abandoning the merging.