Knowledge and Information Systems Взаимодействие человека и компьютера

Self-adaptive improved fungal growth optimization: a hybrid metaheuristic framework for high-quality, robust, and scalable clustering

Das, Kalyan Kumar, Sarkar, Souvik, Majumdar, Anik, Mukhopadhyay, Anirban
Springer Nature · 2026-11-17 · DOI 10.1007/s10115-026-02935-7

Аннотация

Clustering is a fundamental task in machine learning and data science. Traditional clustering algorithms are highly sensitive to initialization and are often prone to becoming trapped in local optima, particularly when dealing with complex or high-dimensional datasets. To overcome these limitations, this article proposes a Self-Adaptive Improved Fungal Growth Optimization (SAIFGO) algorithm, an enhanced version of the nature-inspired Fungal Growth Optimization (FGO) algorithm. The proposed SAIFGO algorithm mimics key fungal behaviors, including hyphal growth, chemotropism, branching, and spore germination. It incorporates several advanced strategies, such as chaotic initialization, Lévy flight-based mutation, opposition-based learning, and an adaptive exploration–exploitation mechanism, to improve the search process. Furthermore, a composite multi-objective fitness function based on multiple cluster validity indices is employed to achieve superior clustering performance. The proposed algorithm is also hybridized with conventional clustering techniques, including K-Means and Fuzzy C-Means (FCM), to optimize cluster centroids more effectively. Extensive experiments are conducted by comparing SAIFGO with several state-of-the-art metaheuristic algorithms using five cluster validity indices and comprehensive statistical analyses. The experimental results demonstrate that SAIFGO consistently outperforms or remains highly competitive with existing methods across diverse datasets. These findings indicate that SAIFGO is an effective, robust, and high-performing optimization algorithm for solving complex data clustering problems.

Читать у издателя →