As we experimented with a reduced feature dataset, where we removed features without a strong correlation to the target variable, we would like to provide the final scores for the best parameters of each method. In the following graph, the results are depicted, sorted in ascending order. Bagging seems to be the most robust method when applied to the filtered dataset. XGBoost is the second best alternative, providing decent F1 and Recall scores when applied to the filtered dataset as well:
F1 scores
Recall scores, depicted in the following plot, show the clear advantage XGBoost has on this metric over the other methods, as it is able to outperform all others for both the original and filtered datasets:
Recall scores
