| 摘要: |
| 【目的】应用以遗传算法优化的BP（GA-BP）神经网络构建灌木生物量估测模型，以有效避免回归分析建模中自变量及模型形式选择的复杂问题。【方法】以灌木林地的荆条为试验对象，应用遗传算法优化BP神经网络的结构、初始权值和阈值，通过BP神经网络训练构建荆条最优地上生物量估测模型，并与传统的应用回归分析方法构建的模型进行对比分析。【结果】仿真结果表明，GA-BP神经网络模型和回归分析模型的模拟精度分别为77.65%和71.79%，估测精度分别为81.46%和75.64%，GA-BP神经网络模型的精度略高于回归分析模型。【结论】应用GA-BP神经网络构建灌木生物量模型是可行的，能够实现灌木生物量的快速估测。 |
| 关键词: 灌木生物量 BP神经网络 遗传算法 回归分析 |
| DOI： |
| 分类号: |
| 基金项目:国家“863”高技术研究发展计划项目（2012AA102001） |
|
| GA-BP neural network based on estimation model of shrub biomass |
|
CAO Shan-shan,SUN Wei,LIU Peng-ju,et al
|
| Abstract: |
| 【Objective】An estimation model of shrub biomass based on the Genetic Algorithm optimized Back Propagation (GA-BP) neural network was built to effectively avoid the complex problems existing in the selection of independent variables and model forms in regression modeling.【Method】The dominant shrub vitex in shrub land was taken as a case study.The structure,initial weights and thresholds of BP neural network were optimized by genetic algorithm,and then the best vitex aboveground biomass estimation model based on the GA-BP neural network (GA-BP-VABEM) was constructed by training the optimized BP neural network.The GA-BP-VABEM was then compared with the RA-VABEM (Vitex Aboveground Biomass Estimation Model based on the method of Regression Analysis).【Result】The simulation accuracies of GA-BP-VABEM and RA VABEM were 77.65% and 71.79% and their prediction accuracies were 81.46% and 75.64%,respectively.Thus,the accuracy of GA-BP-VABEM was slightly higher than that of RA-VABEM.【Conclusion】It is feasible to establish the estimation model of shrub biomass by utilizing GA-BP neural network and the built model has the ability to achieve the rapid estimation of shrub biomass. |
| Key words: shrub biomass BP neural network genetic algorithm regression analysis |
