max_class (comparing max each model instead of each label)
uncertainty_lowest (merge with uncertainty_threshold; same as uncertainty_lowest with default threshold)
uncertainty_threshold
uncertainty_weighted
confidence_weigted
In addition of the ensemble above I changed the print function for debug, warn, and error using logger
change the default name for prediction from pred_df to pred_{MODEL_DESCRIPTION} so we can use the same feature from the same folder with different models and configurations
add random state to SVM and balancing methods so the result is the same for each run
More ensembles:
uncertainty_lowest(merge with uncertainty_threshold; same as uncertainty_lowest with default threshold)In addition of the ensemble above I changed the print function for debug, warn, and error using logger
change the default name for prediction from
pred_df
topred_{MODEL_DESCRIPTION}
so we can use the same feature from the same folder with different models and configurationsadd random state to SVM and balancing methods so the result is the same for each run