改进果蝇优化算法优化广义回归神经网络的预测问题研究

Improved Optimized General Regression Neural Network Based on Fruit Fly Optimization Algorithm for Forecast Problem

为提高传统非线性预测模型的预测精度,提出一种基于改进果蝇优化算法优化广义回归神经网络的预测方法,将果蝇群体分两部分分别进行迭代寻优,从而改进了果蝇优化算法的寻优性能,进而避免了在寻优过程中陷入局部最优。该方法利用改进果蝇优化算法优化广义回归神经网络的径向基函数扩展参数,然后用训练好的广义回归神经网络预测模型进行预测,最后通过订单预测算例进行实证研究。实证研究结果显示,该方法在解决订单预测问题中与未改进的果蝇优化算法优化广义回归神经网络和传统的广义回归神经网络方法对比,具有更高的预测精度和更好的非线性拟合能力。

In order to improve the prediction accuracy of traditional nonlinear prediction model,an improved forecasting method of optimized general regression neural network based on fruit fly optimization algorithm（FOA）was proposed.In this modified FOA algorithm,the fruit flies were divided into two parts iterative optimization to avoid the optimization process into a local optimum,which improved FOA optimizing performance.The modified FOA was used to optimize the speed of radial basis function expansion of generalized regression neural network,and then using the trained general regression neural net-work prediction model to forecast,finally,an order forecasting numerical example is used to empirical research.The empiri-cal results show that the method in solving the order forecasting problems has higher prediction accuracy and better nonlin-ear fitting than general regression neural network based on fruit fly optimization algorithm and traditional generalized regres-sion neural network.