An Improved Firefly Algorithm for Mining High Utility Itemsets

High-Utility Itemset Mining (HUIM) is a pivotal subfield of data mining that focuses on identifying itemsets with high utility rather than merely frequent patterns. Traditional HUIM algorithms are hindered by the exponential complexity of their search space, rendering them inefficient for large-scale databases. Evolutionary Computation (EC)-based approaches have emerged as promising alternatives, offering approximate, yet effective solutions. However, the existing EC-based HUIM methods still encounter considerable computational overhead when extracting true High-Utility Itemsets (HUIs). To address this constraint, this paper proposes a Firefly Algorithm (FA) and a modified FA for HUIM, incorporating a novel flexible inertia weight predicated on logarithmic decrement. This innovation leads to faster convergence and increased mining efficiency. The experimental evaluations on four publicly available benchmark datasets demonstrate that the proposed approach significantly outperforms state-of-the-art algorithms, including HUIM-Genetic Algorithm (GA), Hybrid Ant Colony Optimization (ACO)/GA, and HUIM-FA in terms of utility discovery, execution time, and scalability. The results indicate that the method is highly scalable and maintains robust performance across increasing transaction volumes and item dimensions. It achieves superior runtime, memory efficiency, and a high number of HUIs while sustaining a strong convergence rate, thus establishing its practicality for real-world, large-scale HUIM applications.