def predict_get_spearman_value(test_set, regressor):
test_df = pd.DataFrame(test_set['age_in_days'])
test_df['predicted'] = regressor.predict(
test_set.loc[:, test_set.columns != 'age_in_days'])
spearman_values = use_spearmanr(test_set['age_in_days'].values,
test_df['predicted'].values)
return test_df, spearman_values
def predict_get_spearman_value(X, y, regressor):
predicted_y = regressor.predict(X)
spearman_values = use_spearmanr(y, predicted_y)
pearson_values = use_pearsonr(y, predicted_y)
return predicted_y, spearman_values, pearson_values
def calc_results(y_train_dict,
y_test_dict,
algo_name,
n_rows=3,
n_cols=3,
visualize=False):
train_res = {}
test_res = {}
for sample_num, ((key, train), (_, test)) in enumerate(
zip(y_train_dict.items(), y_test_dict.items())):
if visualize:
if len(y_train_dict.items()) == 1:
n_rows = n_cols = 1
bbox_to_anchor = None
else:
bbox_to_anchor = (0.5, -0.65)
subplot_idx = sample_num % (n_rows * n_cols)
if subplot_idx == 0 and sample_num != 0:
plt.subplots_adjust(hspace=1.1, wspace=0.8)
plt.show()
train_res[key] = {
'spearman':
use_spearmanr(train['y_train_values'],
train['y_train_predicted_values']),
'pearson':
use_pearsonr(train['y_train_values'],
train['y_train_predicted_values'])
}
test_res[key] = {
'spearman':
use_spearmanr(test['y_test_values'],
test['y_test_predicted_values']),
'pearson':
use_pearsonr(test['y_test_values'],
test['y_test_predicted_values'])
}
print('\n**********')
print('Sample number: ', str(sample_num))
print(key)
print('train ' + str(train_res[key]) + ', test' + str(test_res[key]))
if visualize:
spearman_train_data = train_res[key]['spearman']
pearson_train_data = train_res[key]['pearson']
plt.subplot(n_rows, n_cols, subplot_idx + 1)
plt.scatter(train['y_train_values'],
train['y_train_predicted_values'],
label='Train\nSpearman - rho: ' +
str(round(spearman_train_data['rho'], 2)) + ' pval: ' +
str(round(spearman_train_data['pvalue'], 10)) +
'\nPearson - rho: ' +
str(round(pearson_train_data['rho'], 2)) + ' pval: ' +
str(round(pearson_train_data['pvalue'], 10)))
spearman_test_data = test_res[key]['spearman']
pearson_test_data = test_res[key]['pearson']
plt.scatter(test['y_test_values'],
test['y_test_predicted_values'],
label='Test\nSpearman - rho: ' +
str(round(spearman_test_data['rho'], 2)) + ' pval: ' +
str(round(spearman_test_data['pvalue'], 10)) +
'\nPearson - rho: ' +
str(round(pearson_test_data['rho'], 2)) + ' pval: ' +
str(round(pearson_test_data['pvalue'], 10)))
plt.xlabel('Real Values')
plt.ylabel('Predicted Values')
key_as_list = key.split('|')
title = [
key_as_list[idx * 3] + ', ' + key_as_list[(idx + 1) * 3 - 1] +
', ' + key_as_list[(idx + 1) * 3 - 2]
for idx in range(math.floor(len(key_as_list) / 3))
]
if len(key_as_list) % 3 != 0:
title.append(key_as_list[-2] + ', ' + key_as_list[-1])
plt.title(algo_name + '\n' + '\n'.join(title))
plt.legend(loc='lower center', bbox_to_anchor=bbox_to_anchor)
plt.figure()
train_data = [x['spearman']['rho'] for x in train_res.values()]
plt.scatter(list(range(len(train_data))),
train_data,
label='Train',
linewidth=0.3)
test_data = [x['spearman']['rho'] for x in test_res.values()]
plt.scatter(list(range(len(test_data))),
test_data,
label='Test',
linewidth=0.3)
plt.legend()
plt.title(algo_name + ' Spearman' + r'$\rho$ vs params')
plt.xlabel('sample #')
plt.ylabel(r'$\rho$ value')
plt.show()
return train_res, test_res
def calc_results_and_plot(y_train_values,
y_train_predicted_values,
y_test_values,
y_test_predicted_values,
algo_name,
visualize=False,
title=None,
show=True):
train_res = {
'spearman': use_spearmanr(y_train_values, y_train_predicted_values),
'pearson': use_pearsonr(y_train_values, y_train_predicted_values),
'mse': mean_squared_error(y_train_values, y_train_predicted_values)
}
test_res = {
'spearman': use_spearmanr(y_test_values, y_test_predicted_values),
'pearson': use_pearsonr(y_test_values, y_test_predicted_values),
'mse': mean_squared_error(y_test_values, y_test_predicted_values)
}
spearman_train_data = train_res['spearman']
pearson_train_data = train_res['pearson']
mse_train_data = train_res['mse']
spearman_test_data = test_res['spearman']
pearson_test_data = test_res['pearson']
mse_test_data = test_res['mse']
train_label = 'Train\nSpearman - rho: ' + str(
round(spearman_train_data['rho'], 2)) + ' pval: ' + str(
round(spearman_train_data['pvalue'],
10)) + '\nPearson - rho: ' + str(
round(pearson_train_data['rho'], 2)) + ' pval: ' + str(
round(pearson_train_data['pvalue'],
10)) + '\nmse: ' + str(round(
mse_train_data, 2))
test_label = 'Test\nSpearman - rho: ' + str(
round(spearman_test_data['rho'], 2)) + ' pval: ' + str(
round(spearman_test_data['pvalue'],
10)) + '\nPearson - rho: ' + str(
round(pearson_test_data['rho'], 2)) + ' pval: ' + str(
round(pearson_test_data['pvalue'],
10)) + '\nmse: ' + str(round(mse_test_data, 2))
if visualize:
fig = plt.figure()
ax = plt.subplot(111)
ax.scatter(y_train_values, y_train_predicted_values, label=train_label)
ax.scatter(y_test_values, y_test_predicted_values, label=test_label)
plt.xlabel('Real Values')
plt.ylabel('Predicted Values')
title = '' if title is None else title
plt.title(algo_name + '\n' + title)
# ax.axis('equal')
# Shrink current axis's height by 10% on the bottom
box = ax.get_position()
ax.set_position(
[box.x0, box.y0 + box.height * 0.1, box.width, box.height * 0.9])
# Put a legend below current axis
ax.legend(loc='upper center',
bbox_to_anchor=(0.5, -0.05),
fancybox=True,
shadow=True,
ncol=5)
if show:
plt.show()
return train_res, test_res