NeSiFC: Neighbors' Similarity-Based Fuzzy Community Detection Using Modified Local Random Walk

Area of Science:

• Network Science
• Data Mining
• Computational Social Science

Background:

• Community detection is crucial for understanding network structures.
• Existing fuzzy community detection (FCD) methods often require parameter tuning and are sensitive to network characteristics.
• There is a need for robust and adaptable FCD algorithms.

Purpose of the Study:

• To propose a novel neighbors' similarity-based fuzzy community detection (FCD) method called NeSiFC.
• To introduce a modified local random walk (mLRW) and peripheral similarity index (PSI) for enhanced neighbor similarity computation.
• To evaluate the performance of NeSiFC against established FCD algorithms on diverse datasets.

Main Methods:

• NeSiFC computes neighbor similarity using a modified local random walk (mLRW).
• A peripheral similarity index (PSI) is introduced to construct the transition probability matrix for mLRW.
• Fuzzy membership degrees are iteratively calculated using a fuzzy membership function.

Main Results:

• NeSiFC demonstrates outstanding performance in accuracy and quality metrics across various benchmarks and real-world datasets.
• The method effectively detects both disjoint and overlapping community structures.
• Comparative analysis shows NeSiFC outperforms other FCD algorithms like genetic algorithms, membership degree propagation, and center-based fuzzy graph clustering.

Conclusions:

• NeSiFC offers a robust and parameter-light approach to fuzzy community detection.
• The method's independence from network characteristics and minimal parameter tuning make it highly adaptable.
• NeSiFC represents a significant advancement in fuzzy community detection, providing superior results for complex network analysis.