Examples of Questions:
The unprocessed data have three dimensions:
Notes: Decision tree boosting using 100 trees with a depth equal to 5. The trees are estimated with cost parameters \(\lambda={10}^k\) for 100 different \(k\) equidistantly selected from \(\left[-4,\ -1\right]\). The relative influence plot is drawn for the cost parameters \(\lambda\) that minimize the prediction errors.
Notes: Decision tree boosting using 100 trees with a depth equal to 5. The trees are estimated with cost parameters \(\lambda={10}^k\) for 100 different \(k\) equidistantly selected from \(\left[-4,\ -1\right]\). The relative influence plot is drawn for the cost parameters \(\lambda\) that minimize the prediction errors.
Notes: DL models have 3 dense layers with ReLU activations followed by dropout layers. The final activation uses SoftMax. Training obs. = 900. Test obs. = 165. Validation obs. = 225. Trained for 500 epochs.