摘要
以杉木人工林为研究对象,选取与蓄积量预测有关的因子作为样本输入,通过分析神经网络各参数对网络性能的影响得到最佳参数值,构建结构为10∶3∶1的杉木林蓄积量BP神经网络模型,通过模型训练随机抽取46个样本单元数据并预测20个检验样本。结果表明:BP神经网络对于林分蓄积量具有很好的模拟效果,总体拟合精度为88.5%,均方误差MSE=2.95,所构模型合理、稳定,能够快速有效预测杉木林的变化规律。
Taking artificial Chinese fir as study object, this paper selected factors related to stock volume prediction as input samples. The best parameter values were obtained by analyzing the impact of parameters on the neural networks performance, and then the BP neural network model of Chinese fir volume was built with the structure of 10： 3： 1. Finally, 46 samples were selected randomly by model training and 20 test samples were predicted. The results showed that there was a good simulation of stock volume by BP neural networks, with overall fitting accuracy 88.5%, and mean square error 2.95, illustrating that the model was reasonable, stable and able to predict the law of Chinese fir changes fast and effectively.
出处
《福建林学院学报》
CSCD
北大核心
2012年第4期310-315,共6页
Journal of Fujian College of Forestry
基金
福建省林业厅科技研究项目(DH-397)
关键词
杉木林
人工神经网络
BP算法
蓄积量预测
Chinese fir
arcial neural networks
BP algorithm
stock volume prediction
