Community Detection Using Dual-Representation Chemical Reaction Optimization.

Area of Science:

• Network Science
• Computational Intelligence
• Data Mining

Background:

• Complex networks often exhibit community structures, crucial for understanding network organization and function.
• Detecting these community structures is an NP-hard problem, necessitating efficient algorithmic approaches.

Purpose of the Study:

• To propose a novel evolutionary algorithm, dual-representation Chemical Reaction Optimization (DCRO), for effective community detection in complex networks.
• To enhance the balance between exploration and exploitation in evolutionary search for community detection.

Main Methods:

• Developed a dual-representation encoding strategy (locus-based and vector-based) for solutions within the Chemical Reaction Optimization framework.
• Designed two specialized operators tailored to each representation to facilitate diversification and intensification search.
• Applied the proposed DCRO algorithm to both synthetic and real-world network datasets.

Main Results:

• DCRO demonstrated the ability to identify community structures closely resembling actual ones.
• The algorithm achieved performance comparable to several existing state-of-the-art community detection methods.
• The dual-representation approach effectively balanced exploration and exploitation during the search process.

Conclusions:

• DCRO offers a robust and effective approach for community detection in complex networks.
• The proposed dual-representation strategy enhances the performance of Chemical Reaction Optimization for this specific problem.
• DCRO presents a promising alternative to existing methods for uncovering network community structures.