Banca de DEFESA: ALEXANDRE CAMPELO DE CARVALHO

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : ALEXANDRE CAMPELO DE CARVALHO
DATE: 29/05/2023
TIME: 09:00
LOCAL: Video conferência
TITLE:

ARTIFICIAL NEURAL NETWORKS AND LINEAR REGRESSION FOR ESTIMATING VOLUME OF TRUNKS IN FLONA SARACÁ-TAQUERA, STATE OF PARÁ

KEY WORDS:

Logging, Standing trees, Defects in trees, Artificial Intelligence; Natural Forest.

 

PAGES: 52
BIG AREA: Ciências Agrárias
AREA: Recursos Florestais e Engenharia Florestal
SUBÁREA: Manejo Florestal
SUMMARY:

In the Amazon, a major challenge for forest management is estimating the volume of holes in tree trunks, especially in public forests under concession contracts. In these contracts, there is no provision for discounting the hollow in the logs when quantifying the exploited volume and respective payment of the forestry price to the government. The use of models based on machine learning techniques, a subfield of artificial intelligence, is a potential indirect method to find efficient predictive models for estimating hollow volume in commercial tree trunks. Thus, the objective of this research was to evaluate the efficiency of artificial neural networks against regression models to estimate the hollow volume in tree trunks in a natural forest in the Amazon region. The variables of 213 sample trunks were measured, such as: length of the log, diameters of the base and top of the log, diameters of the base and top of the hollow, length of the hollow, DBH and shaft height. To estimate the hollow volumes through machine learning, twelve artificial neural networks (ANN) were trained, and six non-linear regression models were adjusted. Each network had from 2 to 16 neurons in the input layer. For predictive modelling purposes, the original dataset (n = 213) was divided into training data (70%) and test data (30%), using the Leave-One-Out Cross Validation (LOOCV) method to evaluate the performance of the models during the learning process. ANNs of the Multilayer Perceptron (MLP) type were used, with a Resilient Propagation algorithm for weight learning, and a linear activation function. Stopping criteria were established based on mean error and number of cycles (1% error and 30,000 cycles). Regression models were fitted using the Levenberg-Marquardt algorithm. It was observed that the R9 neural network had the lowest estimation error (4.9721%) among the tested networks. While the M1 regression equation presented the smallest estimation error (2.6004%). It is concluded that the R9 network with the input variables hollow length, diameter of the base and top of the hollow was able to accurately describe and estimate the hollow volume of logs from Flona de Saracá - Taquera. However, non-linear regression modelling was even more accurate, using mean hollow diameter and hollow length as independent variables. Both methods were especially useful for estimating hollow volume in felled logs, as it is difficult to accurately determine hollow volume in standing trees.

BANKING MEMBERS:
Presidente - 142.060.489-91 - JOSE NATALINO MACEDO SILVA - Oxford
Interno - 2411802 - RODRIGO GERONI MENDES NASCIMENTO
Externo à Instituição - EDSON MARCOS LEAL SOARES RAMOS - UFPA
Externa à Instituição - GABRIELA CRISTINA COSTA SILVA - UFPA