k-NEAREST NEIGHBOR AND LINEAR REGRESSION IN THE PREDICTION OF THE ARTIFICIAL FORM FACTOR

Abstract

k-nearest neighbor and linear regression in the prediction of the artificial form factor. The proposal of this studywas to test whether the performance of the nonparametric approach k-NearestNeighbor (k-NN), would improve estimatesof individual artificial form factor (f1.3) oftrees of the hybrid Eucalyptus urophylla x Eucalyptus grandiscompared to the Ordinary Least Squares method.A total of149 sample-trees were selected, felled,and diameter was measured along the trunk at10% (d0.1), 30% (d0.3), 50% (d0.5) and 70% (d0.7) of commercial height and posteriorly at 2m intervals. Mathematical models recognized in the literature for predicting the form factor were adjustedfor comparison. The hyperparameter k of optimum adjustment for the k-NN estimator was obtained by repeated cross-validation. The training data of the k-NN regression model were identical to those used in the adjustment of the linear regression models since most multiple linear regression models present problems of collinearity or multicollinearity. The use of the covariate(d0.3.d0.7)/d1.32 and k = 15 made it possible to construct k-NN models with better generalization capacity. The potential of the k-NN estimator to predict the artificial form factor and thus to obtain less biased estimates of individual tree volumes was demonstratedand considered to be superior to the use of linear regression and average form factors. The k-NN approach can be considered more genericforprediction of the tree form factor,and its useis recommended when classical linear regression models or other simpler methods do not yieldgood results.