RISM clustering algorithm based on relative density and inter-cluster connectivity degree.

Area of Science:

• Unsupervised Machine Learning
• Data Mining
• Computational Statistics

Background:

• Clustering complex, non-linear data is challenging, especially determining the optimal number of clusters.
• Existing methods struggle with high dimensionality and intricate data structures.

Purpose of the Study:

• Introduce RISM (Relative Density and Inter-Cluster Connectivity Degree-based Split-and-Merge), a novel clustering algorithm.
• Automatically infer the optimal cluster configuration for complex datasets.
• Improve clustering accuracy, robustness to noise, and scalability.

Main Methods:

• RISM employs a two-phase approach: splitting and merging.
• Splitting phase: Utilizes a novel relative density metric and relative distance to identify subclusters.
• Merging phase: Incorporates inter-cluster distance and connectivity degrees for principled cluster merging.

Main Results:

• RISM demonstrated superior performance compared to nine state-of-the-art algorithms.
• Achieved high clustering accuracy on synthetic and real-world datasets.
• Showed robustness to noise and excellent scalability.

Conclusions:

• RISM effectively addresses the challenge of determining the optimal number of clusters in complex, non-linear data.
• The algorithm's hybrid density- and connectivity-based approach offers significant advantages.
• RISM represents a promising advancement in unsupervised learning for data analysis.