Related to #61. Given a hyperparameter set, analyze the variability of ensemble models due to random undersampling of the majority class. Currently, this also depends on the same random seed- for this task, we must change that.
Expected code:
[ ] A new algorithm, that accepts the undersampling seed as sys arg. In this script, we will:
[ ] Undersample with the sys arg seed
[ ] load the already trained and saved model, which will have certain hyperparameters
[ ] Retrain that model from scratch with our sample
[ ] Save it again with a different name, with reference to the undersampling seed
[x] Comparison scripts. Perhaps we could reuse those from #61.
Related to #61. Given a hyperparameter set, analyze the variability of ensemble models due to random undersampling of the majority class. Currently, this also depends on the same random seed- for this task, we must change that.
Expected code: