基于BP神经网络的土壤适宜性评价——以溪洛渡水电站嘎勒移民安置区为例

人工神经网络具有大规模并行处理、分布式储存、自适应性、容错性等特点,可以解决复杂的非线性问题.本文将BP人工神经网络应用到溪洛渡水电站嘎勒移民安置区土壤适宜性评价中,构建了影响土壤适宜性的评价因子训练集,对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.通过MATLAB神经网络工具箱对专家样本的学习,建立具有泛化能力的土壤适宜性评价BP神经网络模型,确定网络模型结构为9-7-1,均方误差为0.00033,并对预测地块进行评价,得出评价区域以中等适宜性的土壤为主的结果.

Soil Suitability Evaluation Based on BP Neural Network

Artificial neural network with the characteristics of massively parallel processing,distribuion storage,self-adaptive,fault tolerance and etc.,could be used to solve complex nonlinear problems.Therefore,BP artificial neural network was applied for soil suitability assessment,the impact on soil suitability to build evaluation factors,the training set,and the number of hidden layer neurons in the choice of the establishment of the training process and other issues were discussed. Through the neural network study of samples,the generalization ability of neural network model was established to evaluate the prediction block obtained in line with the actual results of the evaluation.