Modified Carnivorous Plant Algorithm Based on Lévy Flight for Optimizing the BP Model

Carnivorous Plant Algorithm (CPA) is a new swarm intelligence optimization algorithm, which simulates the survival and reproduction of carnivorous plants under harsh conditions. Although having the advantages of simple principle and strong optimization ability, CPA is prone to miss the global optimum and fall into slow convergence velocity and low accuracy. Aiming at the shortcomings, a modified carnivorous plant algorithm based on Lévy flight policy (LCPA) is proposed. In the stage of population reproduction and growth, the random number generated by Lévy flight is used to replace the random number generated in the new individual generation formula. The position updated by Lévy flight is compared with the original position, then the better one is retained. Experimental results show that the solution accuracy and convergence speed of LCPA are significantly improved in 12 benchmark test functions, compared with CPA and other 3 common optimization algorithms. Subsequently, the Wilcoxon test is carried out on the experimental results and the test results verify that LCPA has a significant advantage over other algorithms. In the practical application of BP network classification, the applicability of LCPA is further proved.