Figure 5
Average absolute error between observations and predictions for the evaluation data set. At least 50% of the cases are predicted correctly (error equal to zero), while about 30% are predicted incorrectly (error equal to one), with the remainder in between both values. Errors were calculated based on predictions of the evaluation data set, via random forests trained on data with 18% missing values, no outlier or correlated variable removal and followed by imputation of the median value. A 5-fold cross-validation approach was applied and repeated 10 times, with number of individual trees set to 100.
