基于MATLAB6.x的BP人工神经网络的土壤环境质量评价方法研究

对基于MATLAB 6．x的BP人工神经网络工具箱进行了简要的介绍，并将BP人工神经网络应用到土壤环境质量现状评价中。编制了基于MATLAB 6．x土壤环境质量评价程序，并对影响评价结果的训练集的构建、隐层神经元数量的选择、训练过程的建立等问题进行了探讨。结果表明，用随机函数rand或线性函数linspace内插生成网络的训练集是可行的，BP网络隐层的传递函数为tansig。神经元数量为5（用rand函数生成训练集）或8（用linspace函数生成训练集），输出层的传递函数为purelin，神经元数量为1。训练集中加入一定的噪声更有利于提高网络的识别能力。在此基础上，将构建的网络应用到实际土壤环境质量评价中，并将评价的结果与其他评价方法得出的结果进行了比较，表明BP人工神经网络应用到土壤环境质量评价中是切实可行的。

Assessment Method for Soil Environmental Quality Based on BP Neural Network

BP neural network based on MATLAB 6.x and its application in soil environment quality assessment were introduced and the assessment program was built. The influencing factors of BP neural network on soil environmental quality assessment were discussed, which were the input vector, the numbers of neurons in the hidden layers and the training process. Using the interpolation function rand or linspace in building the input vector is feasible. The transfer function of hidden layer is tansig and the numbers of neurons are five or eight when the interpolation function is rand or linspace, while, the transfer function of output layer is purelin and the number of neuron is one. The addition of noise in the input vectors is important to improve the correction of the network. Based on the research results, the constructed network was applied to a case of soil environmental quality assessment and the assessment results were compared with other assessment methods, which suggested that the method of BP network in soil environmental quality assessment was feasible and reasonable.