OLFWA: A novel fireworks algorithm with new explosion operator and two stages information utilization

The Fireworks Algorithm (FWA) represents a relatively new algorithm that has demonstrated its competitiveness among other intelligent optimization algorithms. In most FWA variants, a simple random explosion generates sparks, accompanied by the guiding operator to explore promising search directions. However, this mechanism proves inefficient in exploiting the local landscape and effectively navigating the promising search directions. In this paper, we introduce a novel explosion technique based on orthogonal design (OD), wherein the orthogonal array is leveraged to regulate the distribution of sparks, thus enabling efficient analysis of spark information. Furthermore, we devise a two-stage mechanism for information utilization. In the initial stage, we propose an orthogonal learning (OL) prediction operator to capitalize on local optima by performing factor analysis on both the objective function value and the ranking feature information. Subsequently, in the second stage, an enhanced guiding method is deployed to determine the most promising search direction. The experimental results indicate that the proposed algorithm (OLFWA) showcases competitive performance compared to state-of-the-art FWA variants, various renowned OD-based algorithms, and other promising methods. Moreover, a comprehensive analysis of the effectiveness of each proposed strategy and their interactions has been conducted.