def model_evaluation(y_test_pickle_loc, test_predict_scores_pickle_loc):
"""
"""
## Importing raw entries
y_test = load_df(y_test_pickle_loc)
predicted_scores = load_df(test_predict_scores_pickle_loc)
##
predicted_labels = f_predicted_labels(y_test, predicted_scores, 0.18)
fpr, tpr, thresholds = roc_curve(y_test,
predicted_scores[:, 1],
pos_label=1)
print("\n++ Tabla de referencia: \n{}\n".format(tabla_referencia()))
print("++ Tabla de confusión 1: \n{}\n".format(
tabla_confusion(
confusion_matrix(y_test, predicted_labels, normalize='all'))))
print("++ Tabla de confusión 2: \n{}\n".format(
tabla_confusion(confusion_matrix(y_test, predicted_labels))))
print("\n++ Score accuracy: {}\n".format(
accuracy_score(y_test, predicted_labels)))
precision, recall, thresholds_2 = precision_recall_curve(
y_test, predicted_scores[:, 1], pos_label=1)
thresholds_2 = np.append(thresholds_2, 1)
metrics_report = get_metrics_report(fpr, tpr, thresholds, precision,
recall, thresholds_2)
save_metrics(metrics_report, metrics_report_pickle_loc)
negocio = metrics_report[metrics_report.fpr <= 0.06]
punto_corte = negocio.head(1).threshold.values[0]
new_labels = [
0 if score < punto_corte else 1 for score in predicted_scores[:, 1]
]
print("++ Tabla de confusión reajustada 1: \n{}\n".format(
tabla_confusion(confusion_matrix(y_test, new_labels,
normalize='all'))))
k = 20 / 551
print("\n++ Precision @k: {}".format(
precision_at_k(y_test, predicted_scores[:, 1], k)))
print("\n++ Recall @k: {}\n".format(
recall_at_k(y_test, predicted_scores[:, 1], k)))
df_aeq = pd.DataFrame(y_test)
df_aeq["score"] = new_labels
save_metrics(df_aeq, aequitas_df_pickle_loc)
print("\n** Model evaluation module successfully executed **\n")
def load_transformation(path):
"""
Loading transformation pickle as dataframe for transformation pipeline.
args:
path (string): location where the pickle that will be loaded is.
returns:
df (dataframe): dataframe with features obtained from the transformation module.
"""
df = load_df(path)
return df
def load_model(path):
"""
...
args:
path (string): ...
returns:
-
"""
df = load_df(path)
return df
def load_selected_model_results(path):
"""
...
args:
path (string): ...
returns:
-
"""
df = load_df(path)
return df
def load_features(path):
"""
Loading fe pickle as dataframe from fe pipeline.
args:
path (string): location where the pickle that will be loaded is.
returns:
-
"""
df = load_df(path)
return df
def load_ingestion(path):
"""
Loading ingestion pickle as dataframe for transformation pipeline.
args:
path (string): location where the pickle that will be loaded is.
returns:
df (dataframe): dataframe obtained from ingestion pickle.
"""
df = load_df(path)
return df
def load_transformation(path):
"""
Cargar pickle que se generó durante la transformación
:param path: Path donde se encuentra el pickle
:return:
"""
print("Opening feature engineering pickle from output path")
output_path = os.path.join(path, "output", "fe_df.pkl")
# Recuperar el pickle
incidentes_pkl = load_df(output_path)
print("Feature Engineering pickle successfully retrieved.")
return incidentes_pkl
def load_ingestion(path):
return load_df(path)
def load_transformation(path):
return load_df(path)