Skip to main content
. Author manuscript; available in PMC: 2021 Jun 20.
Published in final edited form as: Sci Total Environ. 2020 Mar 5;722:137661. doi: 10.1016/j.scitotenv.2020.137661

Table 2.

Random Forest Regression Results – Comparing Response Variables

Response Variable R2 Training RMSE Training R2 Test RMSE Test Bias Test R2 Holdout RMSE Holdout Bias Holdout
Conc. GW 0.88 3.6 0.43 7.8 0.20 0.34 10.6 0.34
Conc. SW 0.98 1.7 0.52 3.5 0.16 0.99 1.6 0.13
Percent GW 0.88 7.9 0.28 19.3 0.70 0.17 19.9 3.65
Percent SW 0.35 15.8 0.21 14.6 −0.49 0.40 13.1 3.84

RMSE = Root Mean Squared Error. Training = average of the 10 cross validated model results that used 90% of the balanced dataset for training the model. Test = average of the 10 cross validated models applied on each of the 10% portions of the balanced dataset that was held back when training the models. Holdout = dataset consisting of 20% of the original zero-inflated and unbalanced dataset that was held back when training the models (based on completely separate model runs from the 90/10 training and testing set model runs).