Prognosis of forest production using a neural-fuzzy system and random forest
Name: JÉFERSON PEREIRA MARTINS SILVA
Publication date: 28/02/2018
Advisor:
Name |
Role |
|---|---|
| ADRIANO RIBEIRO DE MENDONÇA | Co-advisor * |
Examining board:
| Name |
Role |
|---|---|
| ADRIANO RIBEIRO DE MENDONÇA | Co advisor * |
| ANTONIO ALMEIDA DE BARROS JUNIOR | External Examiner * |
| GILSON FERNANDES DA SILVA | Internal Examiner * |
Summary: The aim of this study to apply the techniques Random Forest (RF) and Neuro-Fuzzy System (SNF) in the prognosis of forest production. The data used came from continuous forest inventories conducted in settlements of eucalypt clones, located in the south of Bahia. The data processing was performed in Matlab R2016a software. Data were divided into 70% for training and 30% for validation. The algorithms used to generate rules in SNF were Subtractive Clustering (SC) and Fuzzy-C-Means (FCM). The training was done with the hybrid algorithm (descending gradient and least squares) with the number of times varying from 1 to 20. The pertinence functions associated with the input variables were of the gaussian type and the linear output. Several RF were trained by varying the number of trees from 50 to 850 and the number of observations per leaves from 5 to 35. The modeling of the forest production of clonal eucalypt stands can be performed with SNF and RF. The SC and FCM algorithms provide accurate estimates of basal area and volume projection. The RF presented inferior statistics in relation to SNF for prognosis of forest production. Both techniques are good alternatives for the selection of variables used in the modeling of forest production.
Keywordsn: Artificial intelligence, ensemble learning, forest measurement.
