Semi-Supervised Clustering With Constraints of Different Types From Multiple Information Sources

Area of Science:

• Computer Science
• Data Mining
• Machine Learning

Background:

• Semi-supervised clustering enhances cluster analysis using prior knowledge as constraints.
• Existing algorithms often rely on single-source constraints, limiting performance.
• Integrating multi-source constraints is crucial for improving clustering quality.

Purpose of the Study:

• To develop a semi-supervised clustering algorithm capable of integrating multi-source constraints.
• To propose a uniform representation for diverse constraint types.
• To enhance clustering performance and consensus with multiple constraint sources.

Main Methods:

• Analyzing relations among different constraint types.
• Developing a uniform representation for multi-source constraints.
• Constructing an optimization objective model and solution method.

Main Results:

• The proposed algorithm effectively integrates multi-source constraints.
• It reduces the impact of incorrect constraints from single sources.
• Experimental studies demonstrate superior performance compared to existing algorithms.

Conclusions:

• The novel algorithm successfully integrates multi-source constraints for improved clustering.
• This approach enhances cluster structure and consensus, outperforming single-source methods.
• The method offers a robust solution for high-quality semi-supervised clustering.