Multi-view spectral clustering algorithm based on bipartite graph and multi-feature similarity fusion.

Area of Science:

• Machine Learning
• Data Mining
• Computer Science

Background:

• Multi-view clustering faces challenges due to data heterogeneity and inconsistency.
• Existing two-stage spectral clustering methods often result in information loss and suboptimal performance.
• Current fusion strategies struggle with view-specific discrepancies and scalability.

Purpose of the Study:

• To propose a unified multi-view spectral clustering algorithm (BG-MFS).
• To address limitations of existing methods including information loss, view discrepancies, and computational complexity.
• To improve clustering accuracy and efficiency for large datasets.

Main Methods:

• Developed a unified framework (BG-MFS) integrating bipartite graph construction, multi-feature similarity fusion, and discrete clustering.
• Employed a single optimization model for mutual reinforcement of components.
• Introduced an entropy-based weighting mechanism for adaptive view contribution assessment.

Main Results:

• BG-MFS consistently outperforms state-of-the-art methods in clustering accuracy.
• The proposed method demonstrates superior computational efficiency compared to existing approaches.
• Experiments validate the effectiveness of the integrated approach in handling multi-view data.

Conclusions:

• BG-MFS offers a robust and efficient solution for multi-view spectral clustering.
• The unified framework effectively handles data heterogeneity and view-specific discrepancies.
• The algorithm shows significant potential for large-scale multi-view clustering applications.