def prepare_summary_chart(
disparity_df,
metrics_list,
attributes_list,
fairness_threshold,
chart_height,
chart_width,
chart_default_sizes,
):
## If a specific list of attributes was not passed, use all from df
if fairness_threshold is None:
raise ValueError("Fairness threshold cannot be None for the Summary chart.")
metrics = __sanitize_metrics(metrics_list)
if attributes_list is None:
attributes = list(disparity_df["attribute_name"].unique())
else:
attributes = to_list(attributes_list)
__validate_data_inputs(disparity_df, metrics, attributes, fairness_threshold)
# CHART SIZES
(chart_height, chart_width) = __calculate_default_sizes(
chart_height, chart_width, chart_default_sizes, metrics, attributes
)
return (metrics, attributes, chart_height, chart_width)
def attributes(df, attributes):
"""Validates if attributes exists in the dataframe.
:param df: a dataframe from Aequitas
:type df: pandas.core.frame.DataFrame
:param attributes: an attribute or a list of attributes
:type attributes: str or list
"""
attributes = to_list(attributes)
for attribute in attributes:
attribute_df = df[df["attribute_name"] == attribute]
if len(attribute_df) == 0:
raise ValueError(
f"Attribute name {attribute} does not exist in the dataframe.")
def metrics(metrics):
"""Validates if if there are no duplicated metrics, and if all metrics are valid fairness metrics.
:param metrics: a metric or a list of metrics
:type metrics: str or list
"""
metrics = to_list(metrics)
if len(metrics) != len(set(metrics)):
raise ValueError(
f"There is at least one duplicated metric in the metrics list.")
for metric in metrics:
if metric not in METRICS_LIST:
raise ValueError(f"{metric} is not a valid metric.")
def __sanitize_metrics(metrics_list):
try:
return [metric.lower() for metric in to_list(metrics_list)]
except AttributeError:
print('"metrics_list" must be a list of strings.')