Discovering Influential Nodes in Social Networks based on modified independent cascade model and Fuzzy Multi-Objective Genetic Algorithm

چکیده مقاله

The paper presents a novel approach for identifying influential nodes in social networks, addressing the multi-objective nature of the influence maximization problem (IMP). The authors propose a fuzzy-based version of the Non-dominated Sorting Genetic Algorithm II (NSGA-II), referred to as FNSGA, which dynamically adjusts mutation and crossover rates using a fuzzy inference system. This method simultaneously optimizes three objectives: maximizing the spread of influence through a modified independent cascade (IC) model, minimizing the number of initial seed nodes (budget), and minimizing the diffusion time. By replacing computationally intensive Monte Carlo simulations with the Expected Diffusion Value (EDV) metric, the approach significantly reduces execution time while maintaining accuracy.

The key innovation lies in the integration of fuzzy logic to adaptively control the genetic algorithm’s parameters based on real-time feedback, such as the number of chromosomes in the Pareto front and their average fitness. This adaptive mechanism helps prevent premature convergence and enhances the algorithm’s exploration and exploitation capabilities. The proposed method was evaluated on five benchmark social network datasets—Arenasjazz, Canetscience, EgoFacebook, Higgs-Reply, and Slashdot—and compared against several baseline methods, including centrality-based heuristics, non-fuzzy NSGA-II, and recent multi-objective meta-heuristic algorithms such as μGP, MOCSA, GRASP, and MTEF.

Results demonstrate that FNSGA outperforms existing methods across multiple criteria, including higher EDV values, lower seed set costs, greater influence spread, and competitive runtimes. The fuzzy tuning mechanism proved effective in balancing solution quality and computational efficiency, offering a robust and scalable solution for large-scale social networks. The study also highlights the practicality of the approach for real-world applications such as viral marketing, targeted advertising, and crisis management in dynamic social environments.

Overall, this research contributes a flexible and efficient multi-objective framework for influence maximization, combining fuzzy logic with evolutionary computation to address the complex, uncertain, and multi-faceted nature of social network analysis. The authors suggest several future directions, including extending the model to other diffusion frameworks like the linear threshold model, incorporating real-world probabilistic data, and applying the approach to localized or Persian-language social networks for further validation and impact.