Extended Quality (eQual): Radial threshold clustering based on n-ary similarity.

Area of Science:

• Computer Science
• Data Science
• Machine Learning

Background:

• Radial Threshold Clustering (RTC) is a partitioning algorithm grouping similar frames based on seed similarity.
• Existing RTC algorithm exhibits quadratic time complexity (O(n^2)), limiting efficiency with large datasets.
• RTC results are sensitive to input frame order, leading to inconsistent clustering outcomes.

Purpose of the Study:

• To enhance the Radial Threshold Clustering (RTC) algorithm into an improved version named Extended Quality Clustering (EQC).
• To address the scalability and order-dependency issues inherent in the original RTC algorithm.
• To develop a more efficient and robust clustering method for frame analysis.

Main Methods:

• Implemented k-means++ for accelerated seed selection, optimizing the initial seeding process.
• Introduced extended similarity indices to resolve cluster ties, prioritizing densest and most compact clusters.
• Modified the core clustering logic to achieve linear time complexity (O(n)).

Main Results:

• The Extended Quality Clustering (EQC) algorithm achieves linear time complexity, significantly improving efficiency.
• EQC demonstrates order-invariance, producing consistent clustering results regardless of input frame sequence.
• The enhanced algorithm generates more compact and well-separated clusters compared to the original RTC.

Conclusions:

• Extended Quality Clustering (EQC) effectively overcomes the scalability and order-dependency limitations of Radial Threshold Clustering (RTC).
• The novel approach ensures efficient and reliable frame clustering, suitable for large-scale data.
• EQC provides a more robust and accurate clustering solution with improved cluster quality.