def generate_comparing_boxplot(path, metric_one, metric_two):
"""generates a grouped boxplot taking into account two metrics"""
dataset_name = path.split('\\')[-1]
metric_one_path = join(path, metric_one + '.csv')
metric_one_dataset = read_csv_(metric_one_path)
metric_two_path = join(path, metric_two + '.csv')
metric_two_dataset = read_csv_(metric_two_path)
# melt datasets
metric_one_dataset_long = metric_one_dataset.melt(var_name='algorithm', value_name=metric_one)
metric_two_dataset_long = metric_two_dataset.melt(var_name='algorithm', value_name=metric_two)
metric_two_dataset_long = metric_two_dataset_long.drop(metric_two_dataset_long.columns[0], axis=1)
# concatenate datasets
concatenated_metrics = pd.concat([metric_one_dataset_long, metric_two_dataset_long], sort=False, axis=1)
# melt concatenated dataset
melted = pd.melt(concatenated_metrics, id_vars=['algorithm'], value_vars=[metric_one, metric_two], var_name='Metric')
catplot = sns.catplot('algorithm', hue='Metric', y='value', data=melted, kind="box", legend=False, palette="PuBuGn_d",
height=5, aspect=1.75)
catplot.set(xlabel='Algorithms', ylabel="RMSE")
sns.set_context("paper", font_scale=1.5)
sns.set_style("white")
sns.despine(left=True)
catplot.set_xticklabels(['SLM (FLS) Group', 'SLM (OLS) Group', 'SLM (FLS) + TIE/EDV', 'SLM (OLS) + EDV'], rotation=90) # BE CAREFUL, SPECIFIC VALUES
plt.legend(bbox_to_anchor=(1.05, 1), loc=2, borderaxespad=0., title="Metric", frameon=False) # this changes the legend to outside the plot
# plt.figure()
catplot.fig.set_size_inches(11.69, 8.27)
results_folder_path = join(get_results_folder(), dataset_name)
catplot.fig.savefig(join(results_folder_path, metric_one + '__' + metric_two + '.svg'), bbox_inches='tight')
catplot.fig.savefig(join(results_folder_path, metric_one + '__' + metric_two + '.pdf'), bbox_inches='tight')
def generate_boxplot_error(path, metric_name):
"""generates a boxplot for a certain metric"""
dataset_name = path.split('\\')[-1]
to_read = join(path, metric_name + '.csv')
value = read_csv_(to_read)
fig, ax = plt.subplots()
boxplot = sns.boxplot(data=value, palette="PuBuGn_d")
boxplot.set(xlabel='Algorithms', ylabel='RMSE')
sns.set_context("paper", font_scale=1.5)
sns.set_style("white")
sns.despine(left=True)
boxplot.set_xticklabels(['SLM (FLS) Group', 'SLM (OLS) Group', 'SLM (FLS) + TIE/EDV', 'SLM (OLS) + EDV'], rotation=90) # changed from boxplot.get_xticklabels()-BE CAREFUL, SPECIFIC VALUES
fig.set_size_inches(11.69, 8.27)
results_folder_path = join(get_results_folder(), dataset_name)
fig.savefig(join(results_folder_path, metric_name + '.svg'), bbox_inches='tight')
fig.savefig(join(results_folder_path, metric_name + '.pdf'), bbox_inches='tight')
def generate_test_boxplot(path,
algos_list,
metric_name,
labels_list=None,
output_filename=None,
min_max_y=None,
ylabel='RMSE'):
"""generates test boxplot for a given list of algorithms"""
df_list = []
if (metric_name != 'best_results_testing_value'):
for algo in algos_list:
data_set_name = path.split('\\')[-1]
to_read = join(path, metric_name + '.csv')
# TEMP
#===================================================================
# name_file = ALGOS_MAPPING[algo]['name-file']
# to_read = join(path, name_file + '_' + metric_name + '.csv')
#===================================================================
df = read_csv_(to_read)
df = df[[ALGOS_MAPPING[algo]['name-inside-file']]]
df_list.append(df)
value = pd.concat(df_list, axis=1)
elif (metric_name == 'best_results_testing_value'):
for algo in algos_list:
data_set_name = path.split('\\')[-1]
to_read = join(path, metric_name + '.csv')
df = read_csv_(to_read)
name = ALGOS_MAPPING[algo]['name-inside-file']
if name not in df.keys():
if name == 'SLM':
name = 'slm'
elif name == 'MLP':
name = 'mlp'
else:
print('\tUnknown name:', name)
df = df[name]
df_list.append(df)
value = pd.concat(df_list, axis=1)
if labels_list == None:
labels_list = [
ALGOS_MAPPING[algo]['name-inside-file'] for algo in algos_list
]
fig, ax = plt.subplots()
boxplot = sns.boxplot(data=value, palette="PuBuGn_d")
# print(boxplot.get_yaxis().get_view_interval())
if min_max_y != None:
boxplot.get_yaxis().set_view_interval(min_max_y[0], min_max_y[1])
# boxplot.set(xlabel='Algorithms', ylabel='RMSE')
boxplot.set(ylabel=ylabel)
sns.set_context("paper", font_scale=1.5)
sns.set_style("white")
sns.despine(left=True)
# boxplot.set_xticklabels(labels_list, rotation=90) #changed from boxplot.get_xticklabels()-BE CAREFUL, SPECIFIC VALUES
boxplot.set_xticklabels(labels_list)
fig.set_size_inches(11.69, 8.27)
results_folder = get_results_folder()
if not exists(results_folder):
mkdir(results_folder)
results_folder_path = join(results_folder, data_set_name)
if not exists(results_folder_path):
mkdir(results_folder_path)
if output_filename == None:
output_filename = metric_name
fig.savefig(join(results_folder_path, output_filename + '.svg'),
bbox_inches='tight')
fig.savefig(join(results_folder_path, output_filename + '.pdf'),
bbox_inches='tight')