白桦人工林单木生长的人工神经网络模型研究

以东北林业大学帽儿山实验林场白桦人工林为研究对象，采用MATLAB中log-sigmoid型函数（logsig）和线性函数（purelin）为神经元的作用函数，用林分内单木相对直径、林分密度指数、林分地位指数和林分年龄作为输入变量，以单木胸径生长量作为输出变量，构建了4：n：1的单木生长的BP人工神经网络模型。用200组单木生长数据对网络模型进行训练和检验，得最适宜的网络结构为4：3：1，均方误差函数mse=0．00160179，总体拟合精度为96．86％。本模型在充分跟踪样本数据的同时，又保持树木生长方程的规律性，可供同类条件的林分在进行经营设计时进行分析、计算和模拟和预测等使用。

Study on the Artificial Neural Network Model of Individual Tree Growth in the Betula platyphlla Plantation

Taking the Betula platyphlla plantation in the Mao＇ershan Mountains Experimental Forest Farm of Northeast Forestry University as the research object, the log-sigmoid type function logsig and the linear function purelin of MATLAB software were selected as the action function of neurons with the relative diameter of the individual tree, stand density index, stand site index, and stand age inside the stand as the input variables and growth of diameter at breast height of the individual tree as the output variables. Then the 4 ： n ： 1 BP artificial neural network model of individual tree growth was established. Two hundreds groups of individual tree data were used to train and examine the very neural network model. The results showed that the optimum network structure was 4： 3： 1, mean square error was 0.001 601 79, the general fitting accuracy was 96.86%. At the same time of tracking sample data sufficiently, the regularity of tree growth equation was maintained, which can be provided for analysis, calculation, simulation, and prediction when conducting the management design to the stands of similar conditions.