Neural network models of diameter distribution for Pinus massoniana plantations

A BP neural network model of diameter distribution was created by using relative diameters of trees as input variables, and using accumulated frequencies of tree numbers as output variables, in which the model structure was 1:S:1. The expected value distribution of the training data of the neural network model are breast height diameter and diameter distribution, which are collected from a typical 26-year-old Pinus massoniana plantation plot and it is still in growth period. By using the combination of qualitative and quantitative methods, a network model is selected and it is correspond to the discipline of stand diameter distribution and has a high fitting accuracy. The network structure is 1:2:1 and named FR_dnet2. The total accumulated frequency fitting accuracy of the model is 99.5%.Concretely, the accumulated frequency fitting accuracy of diameter class is 93% to 99.9% and the mean value is 99%. While the frequency fitting the accuracy of diameter class is 82% to 99% and the mean value is 95%. The degree of accuracy of the neural network model is very high, and the neural network modeling technology can be applied as an effective modeling technology for diameter distribution of trees.