robotenique

@app.route("/display-snack")
@login_required
def display_snack():
    my_database = mongo[DATABASE]
    my_file = open("snacks/snacks.json", "rb")
    parsed = json.loads(my_file.read().decode('unicode-escape'))
    snacks = parsed
    s = snacks
    Snack.objects.delete()
    for s in snacks:

robotenique

np.random.seed(42)
df = (dataset
               .rename(columns={"Created At": "publication_date",
                                "Text": "tweet",                                
                                "Retweet Count" : "num_retweets",
                                "Username": "username",
                                "Classificacao": "sentiment"})
               .loc[:, ["publication_date", "tweet", "num_retweets",  "username", "sentiment"]])
df["publication_date"] = pd.to_datetime(df["publication_date"],infer_datetime_format=True)
df["sentiment"] = df["sentiment"].replace({"Negativo": 0, "Neutro": np.random.choice([0, 1]), "Positivo": 1})

robotenique

from fklearn.validation.evaluators import auc_evaluator, logloss_evaluator, precision_evaluator, recall_evaluator, \
                                          combined_evaluators, temporal_split_evaluator

def tweet_eval(target_column, prediction_column, time_column):
    eval_args = dict(target_column=target_column,  prediction_column=prediction_column)
    basic_evaluator = combined_evaluators(evaluators=[
        auc_evaluator(**eval_args),
        logloss_evaluator(**eval_args),
        precision_evaluator(**eval_args),
        recall_evaluator(**eval_args)

robotenique

from fklearn.training.classification import nlp_logistic_classification_learner
from fklearn.training.pipeline import build_pipeline
from fklearn.training.utils import log_learner_time

def training_pipeline(text_cols, target_column, vectorizer_params, logistic_params):
    return log_learner_time(
        build_pipeline(
            nlp_logistic_classification_learner(
                text_feature_cols=text_cols,
                target=target_column,

robotenique

# Check docs here https://scikit-learn.org/stable/modules/generated/sklearn.feature_extraction.text.TfidfVectorizer.html
vectorizer_params =  dict(decode_error="replace",
                          lowercase=True,
                          stop_words=nltk.corpus.stopwords.words("portuguese"),
                          ngram_range=(1, 3),
                          strip_accents=None)
# Check docs here https://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html
logistic_params = dict(random_state=42,
                       n_jobs=-1,
                       solver="lbfgs")

robotenique

from fklearn.validation.evaluators import auc_evaluator, logloss_evaluator, precision_evaluator, recall_evaluator, \
                                          combined_evaluators, temporal_split_evaluator

def tweet_eval(target_column, prediction_column, time_column):
    eval_args = dict(target_column=target_column,  prediction_column=prediction_column)
    basic_evaluator = combined_evaluators(evaluators=[
        auc_evaluator(**eval_args),
        logloss_evaluator(**eval_args),
        precision_evaluator(**eval_args),
        recall_evaluator(**eval_args)

robotenique

from fklearn.preprocessing.splitting import space_time_split_dataset

train_set, intime_outspace_hdout, outime_inspace_hdout, outime_outspace_hdout = \
                                        space_time_split_dataset(df,
                                                 train_start_date="2016-12-31",
                                                 train_end_date="2017-01-10",
                                                 holdout_end_date="2017-02-13",
                                                 split_seed=42,
                                                 space_holdout_percentage=0.2,
                                                 space_column="username",

robotenique

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)])