A New Self-Adaptive Hybrid Approach Based on History-Driven Methods for Improving Metaheuristics

Sina Alizadeh, Malek Mouhoub

Abstract: We propose a new hybrid approach, that we call History-driven Particle Swarm Optimization-Simulated Annealing (HdPSO-SA), to improve metaheuristics performance through collaboration and history-driven methods. Collaboration is per-formed using a Self-Adaptive Binary Space Partitioning tree (SA-BSP tree) to partition search space and guide the hybrid frame-work to the most promising sub-region of a given continuous problem to solve. The hybrid framework consists of three phases. In the first phase, the SA - BSP tree is applied in PSO to record essential information, create the landscape of fitness values, and partition the search space during exploration. The second phase consists of a smart controller to learn the SA-BSP maturity condition to balance exploration and exploitation through HdPSO and SA, respectively. The proposed smart controller determines the appropriate step (iteration) for switching from HdPSO to SA. In the third phase, the search space will be limited to only the most promising sub-region. Then, the information of the best solution (fitness value and position) will be given to SA to exploit the limited search space. The proposed HdPSO-SA is compared to several metaheuristics on ten well-known uni-modal and multimodal continuous optimization benchmarks. The results demonstrate the superiority of HdPSO-SA in returning a good quality solution while reducing the execution time.