针对科莫多算法(KMA)在求解复杂函数和高维情况下容易出现早熟收敛的问题,提出了一种改进的变权科莫多优化算法(VWCKMA).首先利用Tent混沌映射产生的序列对科莫多个体位置进行位置初始化,为全局搜索的多样性奠定基础.然后提出可变惯性权重,分别对不同社会等级的科莫多个体的运动进行不同控制,较好地提高了收敛速度.最后利用Tent混沌映射进行局部扰动,使其能够进行更加精确的局部搜索,避免局部最优值.仿真实验表明,在单峰函数和多峰函数求解的标准差和均值中,VWCKMA在收敛精度和收敛速度方面均有很大的提高.针对实际空气污染物PM2.s预测非线性的问题,利用VWCKMA对BP神经网络的权值和阈值进行迭代寻优,基于最优参数的条件下使用BP神经网络对PM2.5进行预测.实验结果表明预测准确率为85.085%,相比单一 BP神经网络预测准确率提高19.85个百分点,体现VWCKMA具有一定的实践应用价值.
The Komodo Mlipir algorithm(KMA)is prone to premature convergence when solving complex functions and high dimensions.Therefore,an improved variable weight Komodo optimization algorithm(VWCKMA)is proposed.Firstly,the sequence generated by Tent chaos mapping is used to initialize the position of Komodo individuals,laying the foundation for the diversity of global search.Secondly,variable inertial weights are proposed to control the movement of Komodo individuals of different social classes respectively,which improves the convergence speed.Finally,the Tent chaotic mapping is used for local perturbation,so that it can conduct more accurate local searches and avoid local optimal values.The simulation experiments show that VWCKMA has greatly improved the convergence accuracy and convergence speed in the standard deviation and mean of solving unimodal functions and multimodal functions.Aiming to the issue of nonlinearity prediction of actual air pollutants PM2.5,VWCKMA is used to iteratively optimize the weights and thresholds of the BP neural network,and the BP neural network is used to predict PM2.5 based on the optimal parameters.The experimental results show that the prediction accuracy is 85.085%,which is 19.85 percentage points higher than the prediction accuracy of BP neural network,indicating that VWCKMA has certain practical application value.