基于改进海鸥优化算法的BP神经网络及其应用

BP neural network based on improved seagull optimization algorithm and its application

针对传统反向传播(BP)神经网络在预测时随机产生的初始权值、阈值影响准确性的问题,提出一种改进的海鸥优化算法(ISOA)对BP神经网络的初始阈值、权值进行寻优。首先,为提高海鸥优化算法(SOA)的收敛精度和跳出局部最优的能力,使用Sine混沌映射初始化种群,引入非线性参数A,在海鸥攻击时引入乘除策略进行扰动,同时在攻击阶段后引入反向学习策略。然后,使用ISOA优化BP神经网络初始权值、阈值,解决对初值敏感和易陷入局部最优的问题。最后,在冻结裂隙砂岩动态冲击试验中进行峰值应力预测,结果表明:对比原始BP、粒子群优化(PSO)-BP和SOA-BP,ISOA优化后的BP神经网络对峰值应力预测精度更高。

Aiming at the problem that the initial weights and threshold randomly generated by traditional back propagation(BP)neural network affect the accuracy of prediction,an improved seagull optimization algorithm(ISOA)is proposed for optimizing of initial thresholds and weights of the BP neural netwoek.Firstly,to improve the convergence precision of seagull optimization algorithm(SOA)and the ability to jump out of local optimum,the population is initialized using Sine chaotic mapping,the nonlinear parameter A is introduced,the multiplication and division strategy is introduced to disturb the seagull attack,and the reverse learning strategy is introduced after the attack phase.Then,the ISOA is used to optimize the initial weights and thresholds of the BP neural netwoek to solve the problem of sensitivity to the initial values and easy to fall into the local optimum.Finally,peak stress prediction is carried out in the dynamic impact test of frozen fractured sandstone.The results show that compared with original BP,PSO-BP and SOA-BP,the BP NN optimized by ISOA has higher precision in peak stress prediction.