ax[0, 0].set_title('ROC Curves')
ax[0, 1].set_title('Sensivity at Lead Time')
ax[-1, 0].set_xlabel('False Positive Rate')
ax[-1, 1].set_xlabel('Lead Time (h)')
plot_path = '/home/alex/mews/images/metrics/'
for f, period in enumerate(periods):
for i, (mews_scorer, data_level) in enumerate(
zip([utils.base, np.median, utils.worst_case],
[False, False, True])):
if mews_scorer.__name__ == 'base':
case, _ = mews.prepare_case_multiple(period=False,
scorer=mews_scorer,
data_level=data_level)
control, _ = mews.prepare_control(period=False,
scorer=mews_scorer,
data_level=data_level)
else:
case, _ = mews.prepare_case_multiple(period=period,
scorer=mews_scorer,
data_level=data_level)
control, _ = mews.prepare_control(period=period,
scorer=mews_scorer,
data_level=data_level)
case_processor = Process(thresholds=thresh,
scorers=case_scorers,
augmenter=augmenter).per_data
from performance.core import mews, metrics
import numpy as np
import matplotlib
matplotlib.use('Agg')
from scipy.stats import wilcoxon
from performance.core import scorers, Process, augmenters
if __name__ == '__main__':
case, _ = mews.prepare_case_multiple()
thresh = np.arange(4, 5) + 3
scorers = [scorers.PosNeg(tmin=0, tmax=np.inf)]
augmenter = augmenters.ShuffleEventMax(num_samples=1000, max_time=1)
random_processor = Process(thresholds=thresh,
scorers=scorers,
augmenter=augmenter).per_data
augment = augmenters.NoAugment()
processor = Process(thresholds=thresh,
scorers=scorers,
augmenter=augmenter).per_data
counts, raw_counts = metrics.run(case, processor)
random_counts, random_raw_counts = metrics.run(case, random_processor)
print(random_counts[0].shape)
x = ((np.array(random_counts[0])[:, :, 0] - counts[0][:, :, 0]) /
case.shape[0]).reshape(-1, )
from performance.core import mews, utils
matplotlib.use('Agg')
from matplotlib import pyplot as plt
path = '/home/alex/mews/images/profiles/'
for data_level in [True, False]:
l = [np.max, utils.mode, np.median, np.mean]
if data_level:
l += [utils.worst_case]
for mews_scorer in l:
print(mews_scorer.__name__)
for period in [1, 2, 4, 8]:
prepared_case, case = mews.prepare_case_multiple(
period=period, scorer=mews_scorer, data_level=data_level)
for i, k in enumerate(np.unique(prepared_case[:, 0])):
cur_data = case[k]
inds = np.where(prepared_case[:, 0] == k)[0]
f, ax = plt.subplots(6, 1, figsize=(20, 20))
ax[0].stem(prepared_case[inds, 1], prepared_case[inds, 2])
ax[0].set_xlim([24, 0])
ax[0].set_title('MEWS Score')
ax[1].set_title('Vital Signs for Calculated MEWS')
for j, (vital, label) in enumerate(
zip([
'BLOOD PRESSURE SYSTOLIC', 'PULSE',
'R CPN GLASGOW COMA SCALE SCORE', 'RESPIRATIONS',
'TEMPERATURE'
], [