tweet_extractor.py

from fklearn.metrics.pd_extractors import *
eval_auc_col =  "auc_evaluator__sentiment"
eval_logloss_col = "logloss_evaluator__sentiment"
eval_precision_col =  "precision_evaluator__sentiment"
eval_recall_col = "recall_evaluator__sentiment"

base_extractor = combined_evaluator_extractor(base_extractors=[evaluator_extractor(evaluator_name=eval_auc_col),
                                                               evaluator_extractor(evaluator_name=eval_logloss_col),
                                                               evaluator_extractor(evaluator_name=eval_precision_col),
                                                               evaluator_extractor(evaluator_name=eval_recall_col)])
# Create a split evaluator based on the publication date available for a df_evaluation dataframe
def create_year_week_extractor(df_evaluation):
    year_week_splits = sorted((k.replace("split_evaluator__publication_date_", "") for k in df_evaluation.keys() if "publication_date" in k))
    return split_evaluator_extractor(split_col="publication_date", split_values=year_week_splits, base_extractor=base_extractor)