A Feature Selection Method Based on Hybrid Natural Inspired Algorithms

Data mining is one of the fastest-growing research domains in the information industry as a result of the wide availability of numerous data. Feature selection is the important preprocessing step to affect the performance of data mining. Ttimum. Besides that, hybrid algorithms maintain the advantages of each single-mode algorithm and avoid its weakness effectively. In this work, we propose a parallel model for feature selection named THDWL, which is a combination of Differential Evolution (DE), Whale Ohe natural-inspired algorithm is one of the most effective methods for feature selection, but there are some limitations in the single-mode algorithm such as slow efficiency and local optimization Algorithm (WOA), and Lightning Attachment Procedure Optimization (LAPO). In THDWL, the operators of these algorithms are independently implemented on their subpopulations and communicate at the end of each iteration to get the global best solution. Comparisons are conducted between single-mode algorithms and THDWL to verify the performance of the proposed method. The simulation results show that THDWL effectively improves the classification accuracy and convergence speed comparing with the single-mode algorithm.