def get_top_vectorizer_dict(dataframe,
metric_name='accuracy',
top_n=1,
classifier=None):
data_filename_list = dataframeutils.get_unique_values_from_dataframe(
dataframe, 'data_filename')
vectorizer_margin_dict = dict({})
for vectorizer in commonconstants.VECTORIZER_LIST:
vectorizer_margin_dict[vectorizer] = 0.0
if classifier is None:
classifier_name_list = dataframeutils.get_unique_values_from_dataframe(
dataframe, 'classifier_name')
else:
classifier_name_list = [classifier]
top_vectorizer_dict = dict({})
for data_filename in data_filename_list:
for classifier_name in classifier_name_list:
sorted_metric_list = get_top_vectorizer_dict_given_data_classifier_metric(
dataframe, data_filename, classifier_name, metric_name)
if len(sorted_metric_list) >= 2:
margin_value = get_margin_value(sorted_metric_list)
vectorizer_margin_dict[
sorted_metric_list[0][0]] = vectorizer_margin_dict[
sorted_metric_list[0][0]] + margin_value
sorted_metric_list = sorted_metric_list[:top_n]
for sorted_metric in sorted_metric_list:
if sorted_metric[0] in top_vectorizer_dict.keys():
top_vectorizer_dict[sorted_metric[
0]] = top_vectorizer_dict[sorted_metric[0]] + 1
else:
top_vectorizer_dict[sorted_metric[0]] = 1
return top_vectorizer_dict, vectorizer_margin_dict
def get_global_metadata_visualization(less_than_value=None,
greater_than_value=None):
dataframe = dataframeutils.get_dataframe_size_filter(
less_than_value=less_than_value, greater_than_value=greater_than_value)
unique_num_rows = dataframeutils.get_unique_values_from_dataframe(
dataframe, column_name='num_rows')
unique_average_sentence_length = dataframeutils.get_unique_values_from_dataframe(
dataframe, column_name='average_sentence_length')
unique_class_imbalance = dataframeutils.get_unique_values_from_dataframe(
dataframe, column_name='imbalance_measure')
unique_class_labels = list(dataframe['num_class_labels'].values)
unique_num_tokens = dataframeutils.get_unique_values_from_dataframe(
dataframe, column_name='num_tokens')
histogram_plot_iterable = [
unique_num_rows, unique_average_sentence_length,
unique_class_imbalance, unique_class_labels, unique_num_tokens
]
title_iterable = [
'Distribution of rows', 'Distribution of Average Sentence Length',
'Distribution of Class Imbalance', 'Distribution of Class Labels',
'Distribution of Tokens'
]
x_label_iterable = [
'Number of Rows', 'Average Sentence Length', 'Class Imbalance',
'Number of Class Labels', 'Number of Tokens'
]
y_label_iterable = ['Count'] * len(x_label_iterable)
for plot_iter, title_iter, x_label_iter, y_label_iter in zip(
histogram_plot_iterable, title_iterable, x_label_iterable,
y_label_iterable):
filename_to_save = get_filename_to_save([str(title_iter)])
visualizationutils.plot_histogram_chart(
plot_iter,
title_name=title_iter,
x_label=x_label_iter,
y_label=y_label_iter,
filename_to_save=filename_to_save)
def get_top_classifier_dict(dataframe,
metric_name='accuracy',
top_n=1,
vectorizer=None):
data_filename_list = dataframeutils.get_unique_values_from_dataframe(
dataframe, 'data_filename')
if vectorizer is None:
vectorizer_name_list = dataframeutils.get_unique_values_from_dataframe(
dataframe, 'vectorizer')
else:
vectorizer_name_list = [vectorizer]
top_classifier_dict = dict({})
for data_filename in data_filename_list:
for vectorizer_name in vectorizer_name_list:
sorted_metric_list = get_top_classifier_dict_given_data_vectorizer_metric(
dataframe, data_filename, vectorizer_name, metric_name)
sorted_metric_list = sorted_metric_list[:top_n]
for sorted_metric in sorted_metric_list:
if sorted_metric[0] in top_classifier_dict.keys():
top_classifier_dict[sorted_metric[
0]] = top_classifier_dict[sorted_metric[0]] + 1
else:
top_classifier_dict[sorted_metric[0]] = 1
return top_classifier_dict
def get_top_vectorizer_given_category(dataframe,
metric_name='accuracy',
top_n=1,
category_name_list=None,
classifier=None):
category_names = dataframeutils.get_unique_values_from_dataframe(
dataframe, 'category_folder_name')
if category_name_list is not None:
category_names = filter(lambda x: x in category_name_list,
category_names)
category_top_vectorizer_dict = dict({})
for category_name in category_names:
filtered_dataframe = dataframe[dataframe['category_folder_name'] ==
category_name]
top_vectorizer_dict = get_top_vectorizer_dict(filtered_dataframe,
metric_name=metric_name,
top_n=top_n,
classifier=classifier)
category_top_vectorizer_dict[category_name] = top_vectorizer_dict
return category_top_vectorizer_dict