Contextual Correlation Preserving Multiview Featured Graph Clustering.

Area of Science:

• Data Mining
• Graph Analytics
• Machine Learning

Background:

• Graph clustering is vital for analyzing interconnected data, with applications in social networks and biology.
• Existing methods struggle with multi-view features and contextual correlations between data points.
• The increasing volume of graph data necessitates advanced clustering techniques.

Purpose of the Study:

• To develop a novel graph clustering model that addresses limitations of existing methods.
• To effectively utilize multi-view vertex features and contextual correlations for improved clustering.
• To introduce the Contextual Correlation Preserving Multi-View Featured Graph Clustering (CCPMVFGC) model.

Main Methods:

• CCPMVFGC learns a shared latent space from multi-view features to represent cluster preferences.
• It models inter-relationships between vertices using this latent space and computed contextual correlations.
• A unified objective function and iterative strategy are used for optimization, with theoretical analysis provided.

Main Results:

• CCPMVFGC was compared against classical and state-of-the-art methods on eight graph datasets.
• The model achieved competitive performance on both multi-view and single-view datasets.
• Experimental results validate the effectiveness of CCPMVFGC in graph clustering.

Conclusions:

• The proposed CCPMVFGC model effectively integrates multi-view features, contextual correlations, and graph topology for superior clustering.
• The model demonstrates robust performance across diverse graph datasets, highlighting its practical applicability.
• CCPMVFGC offers a significant advancement in graph clustering for complex, multi-faceted data.