PERFORMANCE ANALYSIS OF MACHINE LEARNING TECHNIQUES IN THE CLASSIFICATION OF DAMAGE CAUSED BY CAPUCHIN MONKEYS IN PINE PLANTATIONS

The breaking of paradigms has changed the dynamics of modern forestry, making it multidisciplinary, including continuous improvement processes through natural or artificial methods, such as genetic modification or the introduction of technological tools. In this sense, the objective of this work is to draw profiles on the pattern of damage caused by Sapajus nigritus Goldfuss, 1809, popularly known as capuchin monkey, in plantations of Pinus taeda L., with the help of machine learning techniques, focusing on the classification of trees that will or will not be attacked. The indicators showed a good convergence of the XGBoost and Random Forest methods, providing 73.5% and 72.4% accuracy, respectively, optimizing their hyperparameters. At the same time, the graphs show the importance of the variables: date of the inventory, confirming the temporal dependence of the data, and height and strip, which suggests a possible geographical relationship between the action of capuchin monkeys in areas far from roads or facilities.