We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Uso de regressão e redes neurais artificiais na estimativa do volume de Khaya ivorensis.
- Authors
Sousa Lopes, Lucas Sérgio; Rode, Rafael; Pauletto, Daniela; Damázio Baloneque, Diego; dos Santos, Fábio Guerra; Resende Silva, Arystides; Breda Binoti, Daniel Henrique; Garcia Leite, Helio
- Abstract
This work aimed to compare the volume of African mahogany trees estimated by the Schumacher and Hall model and artificial neural networks (ANNs). Data collection took place in two agroforestry systems in the municipality of Belterra, Pará, Brazil, with 7 and 11 years of age. At each site 34 standing trees were planted. For commercial volume estimates, the forms of the Schumacher and Hall model (linear and nonlinear) and the use of artificial neural networks (RNA) of the Multilayers perceptron type were used. The ANNs architectures with 4 neurons in the input layer provided the best estimates and error values, which are appreciably better than the volumetric models, with RNA being 36.7% smaller than the non-linear Schumacher and Hall model. The latter model tended to overestimate volumes and RNA obtained more trend-free estimates. Artificial neural networks generated estimates more accurately than the regression model. This technique proved to be feasible, since a single network can estimate the volume for different locations, avoiding the need for stratification.
- Publication
Brazilian Journal of Wood Science / Revista Ciência da Madeira, 2020, Vol 11, Issue 2, p74
- ISSN
2177-6830
- Publication type
Article
- DOI
10.12953/2177-6830/rcm.v11n2p74-84