林分保留密度的人工神经网络模型研究

为探索人工神经网络在林分保留密度研究中的应用效果,以马尾松人工林为对象,481块标准地资料为基础数据,用MATLAB的正切“S”形函数tansig为神经元的作用函数,以年龄和密度的标准化、归一化数据为训练样本,构建和训练林分保留密度的人工神经网络模型,并与幂函数进行对比研究.结果表明,密度与年龄的关系用幂函数(y=axb)描述,拟合精度为91.29%;用人工神经网络描述,拟合精度为94.27%.可见,人工神经网络能比常规方法更好地模拟林分密度随年龄的变化规律,特别能有效地模拟经营活动下的复杂过程.

Study on stand's reserved density model with ANN.

A stand's reserved density model for Masson pine(Pinus massoniana) plantation was developed with ANN(artificial neural network).The data for the study were collected from 481 sample plots.Using tangent "S" functions of MATLAB as the neural functions,a reserved density model with ANN was established and trained by the trained samples of standardized and mormalized data of age and density.A comparative study between ANN model and power function was also conducted.The fitting accuracy of ANN model was(94.27%),while the fitting accuracy of power function was 91.29%.It is concluded that the ANN model is better than the common methods in simulating the density changing along with age,especially in simulating the complex process of forest management.