def test_add_binning_cols__works(self):
data = [[0.3,],
[0.6,],
[1.0,],
[1.3,],
[1.6,],
[2.0,],
]
input = pd.DataFrame(data, columns = ['val'])
result = add_binning_cols(input, prob_col='val', prefix='pre', bins=list(drange_inc(0, 2, '1')), bin_labels=list(range(1, 3)))
# print('result')
# print(result)
item = result.iloc[0]['pre_range']
# print('item')
# print(item)
# print('type')
# print(type(item).__name__)
data = [[0.3, 1, pd.Interval(0.0, 1.0)],
[0.6, 1, pd.Interval(0.0, 1.0)],
[1.0, 1, pd.Interval(0.0, 1.0)],
[1.3, 2, pd.Interval(1.0, 2.0)],
[1.6, 2, pd.Interval(1.0, 2.0)],
[2.0, 2, pd.Interval(1.0, 2.0)],
]
expected = pd.DataFrame(data, columns = ['val', 'pre_ind', 'pre_range'])
assert_array_equal(result, expected)
def test_drange_inc__works(self):
result = list(drange_inc(0, .1, '0.05'))
expected = [0, 0.05, 0.1]
assert_equal(result, expected)
history_ids_validate, pfa_pred_validate, pfa_dash_pred_validate,
dnn_pred_validate, dnn_dash_pred_validate
],
axis=1)
pfa_vs_dnn.to_csv(os.path.join(result_dir,
f'pfa_pred_vs_dnn_pred_w_dash_validate.csv'),
index=False)
# Compute correlations
pfa_vs_dnn_just_pred = pfa_vs_dnn.loc[:, [
'pfa_pred', 'pfa_d_pred', 'dnn_pred', 'dnn_d_pred'
]]
pfa_vs_dnn_just_pred.corr()
bins = list(drange_inc(0, 1, '0.05')) # 5% point bin size
bin_labels = list(range(1, 21))
base_cols = ['pfa', 'pfa_d', 'dnn', 'dnn_d']
correct_cols = [c + "_cor" for c in base_cols]
prediction_cols = [c + "_pred" for c in base_cols]
# correct_cols = ['pfa_cor', 'dnn_d_cor', 'dnn_cor', 'pfa_d_cor']
# prediction_cols = ['pfa_pred', 'dnn_d_pred', 'dnn_pred', 'pfa_d_pred']
pfa_vs_dnn_binned = pfa_vs_dnn.copy()
pfa_vs_dnn_binned = pfa_vs_dnn_binned.drop(columns=correct_cols)
# pfa_vs_dnn_binned.to_csv(os.path.join(result_dir, f'pfa_pred_vs_dnn_pred_w_dash_validate_no_cor.csv'), index=False)
for prob_col in prediction_cols:
add_binning_cols(pfa_vs_dnn_binned,