def make_factors(self, dataframe, number_top_features=3):
"""
Given a prediction dataframe, build and return a list of the top k features in dataframe format.
Args:
dataframe (pandas.core.frame.DataFrame): Raw prediction dataframe
number_top_features (int): Number of top features per row
Returns:
pandas.core.frame.DataFrame: A dataframe containing the grain id and factors
"""
# Run the raw dataframe through the preparation process
prepared_dataframe = self.prepare_and_subset(dataframe)
# Create a new dataframe with the grain column from the original dataframe
results = dataframe[[self.grain_column]]
# Create a list of column names
reason_col_names = ['Factor{}TXT'.format(i) for i in range(1, number_top_features + 1)]
# Get a 2 dimensional list of all the factors
top_features = hcai_factors.top_k_features(prepared_dataframe, self.feature_model, k=number_top_features)
# Verify that the number of factors matches the number of rows in the original dataframe.
if len(top_features) != len(dataframe):
raise HealthcareAIError('Warning! The number of predictions does not match the number of rows.')
# Create a dataframe from the column names and top features
reasons_df = pd.DataFrame(top_features, columns=reason_col_names, index=dataframe.index)
# Join the top features and results dataframes
results = pd.concat([results, reasons_df], axis=1, join_axes=[dataframe.index])
# results.set_index(keys=self.grain_column, inplace=True)
return results
def make_factors(self, dataframe, number_top_features=3):
"""
Given a prediction dataframe, build and return a list of the top k features in dataframe format.
Args:
dataframe (pandas.core.frame.DataFrame): Raw prediction dataframe
number_top_features (int): Number of top features per row
Returns:
pandas.core.frame.DataFrame: A dataframe containing the grain id and factors
"""
# Run the raw dataframe through the preparation process
prepared_dataframe = self.prepare_and_subset(dataframe)
# Create a new dataframe. If grain column exists, add the grain
# column from the original dataframe; otherwise,
# just create a new empty dataframe.
if self.grain_column is not None:
results = dataframe[[self.grain_column]]
else:
results = pd.DataFrame()
# Create a list of column names
reason_col_names = ['Factor{}TXT'.format(i) for i in range(1, number_top_features + 1)]
# Get a 2 dimensional list of all the factors
top_features = hcai_factors.top_k_features(prepared_dataframe, self.feature_model, k=number_top_features)
# Verify that the number of factors matches the number of rows in the original dataframe.
if len(top_features) != len(dataframe):
raise HealthcareAIError('Warning! The number of predictions does not match the number of rows.')
# Create a dataframe from the column names and top features
reasons_df = pd.DataFrame(top_features, columns=reason_col_names, index=dataframe.index)
# Join the top features and results dataframes
results = pd.concat([results, reasons_df], axis=1, join_axes=[dataframe.index])
# results.set_index(keys=self.grain_column, inplace=True)
return results