Improved multi-objective clustering algorithm using particle swarm optimization.

Area of Science:

• Computer Science
• Artificial Intelligence
• Data Mining

Background:

• Multi-objective clustering is gaining attention for improved accuracy.
• Existing methods may face challenges with local optima and efficiency.

Purpose of the Study:

• To propose an improved multi-objective clustering framework using particle swarm optimization (IMCPSO).
• To enhance the search for optimal clustering solutions in continuous space.
• To improve algorithm efficiency and avoid local optima.

Main Methods:

• Designed a novel particle representation for clustering within particle swarm optimization (PSO).
• Analyzed Pareto set distribution to inform leader selection strategies.
• Developed a clustering solution-improvement method to boost search efficiency.

Main Results:

• The proposed IMCPSO framework demonstrated superior performance across 28 datasets.
• Outperformed nine state-of-the-art clustering algorithms based on the ARI evaluation index.
• The novel methods effectively addressed local optima and improved search efficiency.

Conclusions:

• The IMCPSO framework offers a more accurate and efficient approach to multi-objective clustering.
• The novel particle representation and leader selection strategy are key to the algorithm's success.
• This research contributes a robust method for complex clustering problems.