|
Abstract:
|
Nonparametric, multivariate nearest neighbors techniques have been shown to be useful for simultaneously predicting multiple forest attributes from forest inventory and satellite image data. A two-step nearest neighbors algorithm is proposed in which the class of a relevant categorical variable such as land cover is predicted in the first step, and continuous variables such as tree volume are predicted in the second step subject to the constraint that all nearest neighbors must come from the predicted class of the categorical variable. Nearest neighbors, multinomial logistic regression, and discriminant analysis techniques are compared for use in the first step, whereas only the k-Nearest Neighbors (k-NN) technique are used in the second step because most forest inventory variables do not satisfy the distributional assumptions necessary for parametric multivariate techniques.
|
