The three predictive models used to simulate the non-linear sediment-environment
relationships are Boosted Decision Tree (BDT) (Friedman 1999), Support Vector
Machine (SVM) (Cortes and Vapnik 1995), and General Regression Neural Network
(GRNN) (Specht 1990). The models’ performance was evaluated against a separated test
set (78 out of 259 samples) using three statistics. They include R
2
(proportion of variance
explained by model), Root Mean Squared Error (RMSE), and Mean Absolute Error
(MAE). To investigate the sensitivity ofmodel performance to the number of
explanatory variables, the secondary variables were added to the models one at a time
until the best performance was reached.