BP神经网络在数学建模中的应用.docVIP

装订示例 PAGE 19 黔南民族师范学院毕业论文（设计） 参考文献 BP神经网络在数学建模中的应用 摘要：BP神经网络的应用问题是历年来科学家致力解决的问题之一，也是工程中的常见问题，特别在数学建模中的应用，数年以来多方面的探究提出了一系列关于BP神经网络的数值算法。目前，有学者利用人工神经网络具有很强映射能力的特点，运用传统的数值算法来解决许多实际问题，但是由于已有的算法大多存在收敛速度慢，记忆不稳定性突出，网络的中间层和它的单元数选取无理论指导等问题，因此迫切需要寻找更为有效的智能算法。本文主要研究BP神经网络在数学建模中关于函数逼近与短期预测的应用问题。为了改进传统的数值算法，我们通过对改变函数参数以及隐层神经元个数训练结果的探究，分别提出了基于非线性函数的逼近算法和短期预测问题的改进算法。结合函数逼近与虫害预测实例，根据一系列的测试和仿真，最终得出的逼近效果比改进前更为明显，改进后的预测结果与目标值大体相符。所提算法可行性强，操作简便，具有较高的执行效率和良好的安全性。 关键词：BP神经网络；函数逼近；短期预测 The BP neural network in the application of mathematical modeling （mathematics department, Qiannan Normal College for Nationalities, Duyun 558000, Guizhou） Abstract : BP neural network is applied to solve the problem is over the problem of scientists is one of the project of the common problems, especially in the application of mathematical modeling, the years of various explore proposed a series about the BP neural network of numerical algorithm. At present, some scholars using artificial neural network has very strong mapping capability characteristics, using traditional numerical algorithm is proposed to solve many problems, but because most of the existing algorithm existing slow convergence speed, memory not stability is outstanding, the network interface layer and its unit number without theory guidance for selection, so desperate for more effective intelligent algorithm. This paper mainly studies the BP neural network in the mathematical modeling of the function approximation and the short-term application problems. In order to improve the traditional numerical algorithm by us, to change the function parameters and the number of neurons in hidden layer training of exploration results are put forward, based on the nonlinear function approximation algorithm and the short-term forecast problems of the algorithm. Combined with the function approximation and pests forecasting examples, according to a series of tests and simulation, eventually