def test_check_settings_plot_intervals_basic(self):
data = DataStructure()
data.add_data_set(x=np.random.random_sample((100, 2)), y=None)
results = gf.setup_pseudo_results()
chain = results['chain']
s2chain = results['s2chain']
intervals = uqp.calculate_intervals(
chain, results, data, model3D, s2chain=s2chain)
fig, ax, isets = uqp.plot_3d_intervals(
intervals, data.xdata[0], limits=[95],
return_settings=True)
self.assertEqual(ax.get_legend_handles_labels()[1],
['Model', '95% PI', '95% CI'],
msg=str('Strings should match: {}'.format(
ax.get_legend_handles_labels()[1])))
self.assertTrue(isinstance(isets, dict),
msg='Expect dictionary')
self.assertEqual(isets['ciset']['limits'], [95])
self.assertEqual(isets['piset']['limits'], [95])
plt.close()
plt.close()
fig, ax, isets = uqp.plot_3d_intervals(
intervals, data.xdata[0], limits=[95],
adddata=True, ydata=data.xdata[0][:, 0],
return_settings=True)
self.assertEqual(ax.get_legend_handles_labels()[1],
['Model', 'Data', '95% PI', '95% CI'],
msg=str('Strings should match: {}'.format(
ax.get_legend_handles_labels()[1])))
self.assertTrue(isinstance(isets, dict),
msg='Expect dictionary')
self.assertEqual(isets['ciset']['limits'], [95])
self.assertEqual(isets['piset']['limits'], [95])
plt.close()
def test_predintcreation(self):
data = DataStructure()
data.add_data_set(x=np.linspace(0, 1), y=None)
results = gf.setup_pseudo_results()
chain = results['chain']
s2chain = results['s2chain']
intervals = uqp.calculate_intervals(
chain, results, data, model, s2chain=s2chain)
self.assertTrue('credible' in intervals.keys(),
msg='Expect credible intervals')
self.assertTrue('prediction' in intervals.keys(),
msg='Expect prediction intervals')
self.assertTrue(isinstance(intervals['credible'], np.ndarray),
msg='Expect numpy array')
self.assertTrue(isinstance(intervals['prediction'], np.ndarray),
msg='Expect numpy array')
intervals = uqp.calculate_intervals(
chain, results, data, model, s2chain=0.1)
self.assertTrue('credible' in intervals.keys(),
msg='Expect credible intervals')
self.assertTrue('prediction' in intervals.keys(),
msg='Expect prediction intervals')
self.assertTrue(isinstance(intervals['credible'], np.ndarray),
msg='Expect numpy array')
self.assertTrue(isinstance(intervals['prediction'], np.ndarray),
msg='Expect numpy array')
def test_plot_intervals_basic(self):
data = DataStructure()
data.add_data_set(x=np.linspace(0, 1), y=None)
results = gf.setup_pseudo_results()
chain = results['chain']
s2chain = results['s2chain']
intervals = uqp.calculate_intervals(
chain, results, data, model, s2chain=s2chain)
fig, ax = uqp.plot_intervals(intervals, data.xdata[0], limits=[95])
self.assertEqual(ax.get_legend_handles_labels()[1],
['Model', '95% PI', '95% CI'],
msg=str('Strings should match: {}'.format(
ax.get_legend_handles_labels()[1])))
plt.close()
fig, ax = uqp.plot_intervals(intervals, data.xdata[0], limits=[95],
adddata=True, ydata=data.xdata[0])
self.assertEqual(ax.get_legend_handles_labels()[1],
['Model', 'Data', '95% PI', '95% CI'],
msg=str('Strings should match: {}'.format(
ax.get_legend_handles_labels()[1])))
plt.close()
def main():
#############################################################
print('Getting data from github...')
try:
subprocess.run([
'gitdir',
'https://github.com/CSSEGISandData/COVID-19/tree/master/csse_covid_19_data/csse_covid_19_daily_reports/'
])
directory = './csse_covid_19_data/csse_covid_19_daily_reports'
df = dfProcess(directory, 'US')
except:
print('Directory error.')
print('Countries:\n {} \n'.format(df.countries))
print('Dates in database:\n {} \n'.format(df.dates))
print('Data for US available from the following dates:\n {} \n'.format(
df.cDates))
print('Data for US States Isolated: \n {}'.format(df.states))
############################################################
'''Map Creation'''
#Will take a while
#Create maps for all dates
print('\n Creating maps for each available date...')
mpl.use('Agg')
for i, date in enumerate(df.country.keys()):
print("Creating map for {}.........{}/{}".format(
date, i + 1, len(df.country.keys())))
df.mapPlot(df.country[date], 'Deaths', 1)
print('Maps are saved in Progress folder...')
############################################################
'''Stitching images together'''
print('Creating video of maps...')
path = sorted(glob.glob(os.path.join('./Progress/*.png')))
files = [cv2.imread(file) for file in path]
height, width, layers = files[0].shape
size = (width, height)
files = [cv2.resize(file, size) for file in files]
out = cv2.VideoWriter('ConfirmedvsDeaths.avi',
cv2.VideoWriter_fourcc(*'XVID'), 60, size)
for file in files:
for _ in range(15):
out.write(file)
out.release()
print('Done.')
#########################################################
'''Parameter estimation'''
#Define cost function for MCMC
def ssfunc(q, data):
y = data.ydata[0]
beta, gamma, N, S, I, R = q
ymodel = model.runmodel(beta, gamma, N, S, I, R)
res = ymodel[1] - y
return (res**2).sum()
print('Setting up auxilary files and MCMC parameters')
#Aux data
census = pd.read_excel('nst-est2019-01.xlsx', encoding='UTF-8')
census['State'] = list(
map(lambda x: re.sub(r'\.', '', x), census['State']
)) #Some states has a period in the beginning, cleaning that
census.set_index('State', inplace=True)
#Scatter
confirmed = [
j for k in df.states.columns.levels[0][:-2]
for j in df.states[k]['Confirmed']
]
timespan = np.linspace(0, len(df.cDates), len(confirmed))
model = SIR(timespan, 14)
#invoke module
mcmcstat = MCMC()
mcdata = DataStructure(
) #Just doing this for the sake of following proper format
mcmcstat.data.add_data_set(x=timespan, y=confirmed)
#apply cost function
mcmcstat.model_settings.define_model_settings(sos_function=ssfunc)
#simulation options. Using DRAM method
mcmcstat.simulation_options.define_simulation_options(nsimu=10.0e3,
updatesigma=True,
method='dram',
adaptint=100,
verbosity=1,
waitbar=1)
#Parameters of interest
mcmcstat.parameters.add_model_parameter(name='beta',
theta0=0.01,
minimum=0.00001,
maximum=0.5)
mcmcstat.parameters.add_model_parameter(name='gamma',
theta0=0.01,
minimum=0.001,
maximum=0.5 / 2.1)
'''Population and initial conditions, these are fixed.'''
N, S, I = census['Population'].sum() / len(
df.states), census['Population'].sum() / len(df.states) - 1, 1
R = 0
mcmcstat.parameters.add_model_parameter(name='N', theta0=N, sample=False)
mcmcstat.parameters.add_model_parameter(name='S', theta0=S, sample=False)
mcmcstat.parameters.add_model_parameter(name='I', theta0=I, sample=False)
mcmcstat.parameters.add_model_parameter(name='R', theta0=R, sample=False)
print('Running MCMC. Will take a while...')
mcmcstat.run_simulation()
print('Done.')
####MCMC outputs
mcmcresults = mcmcstat.simulation_results.results
burnin = int(mcmcresults['nsimu'] / 2)
#Statistics
chain = mcmcresults['chain']
s2chain = mcmcresults['s2chain']
sschain = mcmcresults['sschain']
names = mcmcresults['names']
mcmcstat.chainstats(chain[burnin:, :], mcmcresults)
print('Acceptance rate: {:6.4}%'.format(
100 * (1 - mcmcresults['total_rejected'])))
print('Model Evaluations: {}'.format(mcmcresults['nsimu'] -
mcmcresults['iacce'][0] +
mcmcresults['nsimu']))
if not os.path.exists('MCMC'):
os.makedirs('MCMC')
fig = mcp.plot_density_panel(chain[burnin:, :], names)
fig.savefig("MCMC/Density.png", bbox_inches='tight', dpi=100)
fig = mcp.plot_pairwise_correlation_panel(chain[burnin:, :], names)
fig.savefig("MCMC/Pairwise.png", bbox_inches='tight', dpi=100)
print('MCMC parameter plots are saved in MCMC')
##########################################################
print('Creating final projection graph...')
#difference from first date to last day
ddiff = abs((dt.strptime(df.cDates[0], '%m-%d-%Y') -
dt.strptime(df.cDates[-1], '%m-%d-%Y')).days)
#projected parameters
pbeta, pgamma = mcmcresults['mean'][0], mcmcresults['mean'][1]
#Projecting 360 days, fingers crossed
ptimespan = np.linspace(ddiff, ddiff + 365, 365 * len(df.states))
model = SIR(ptimespan, 14)
#define current susceptible, infected, removed.
cconfirmed = df.statsPerCountry('US').loc[
df.cDates[-1]]['Confirmed'] / len(df.states)
cremoved = (df.statsPerCountry('US').loc[df.cDates[-1]]['Recovered'] +
df.statsPerCountry('US').loc[df.cDates[-1]]['Deaths']) / len(
df.states)
csus = S - cconfirmed - cremoved
projection = model.runmodel(pbeta, pgamma, N, csus, cconfirmed, cremoved)
pfig, pax = plt.subplots(1, 1, figsize=(35, 8))
pax.plot(ptimespan, projection[1])
pax.set(title='Projected Infectives', xlabel='Days since 03-01-2020')
plt.savefig('./FinalProjection.png', bbox_inches='tight', dpi=100)
print('FinalProjection graph saved in current folder...Done')