weights = [1]
# set PSO optimisation settings
pars = ['warmup', 'beta', 'l', 'effectivity']
bounds = ((20, 40), (0.01, 0.09), (0.1, 20), (0.03, 0.97))
# run PSO optimisation
theta = pso.fit_pso(model,
data,
pars,
states,
bounds,
maxiter=maxiter,
popsize=popsize,
start_date=start_calibration)
theta = np.array([37.45480627, 0.04796753, 11, 0.14]) #-75918.16606140955
# Assign estimate
warmup, model.parameters = assign_PSO(model.parameters, pars, theta)
# Perform simulation
out = model.sim(end_calibration, start_date=start_calibration, warmup=warmup)
# Visualize fit
ax = plot_PSO(out, theta, pars, data, states, start_calibration,
end_calibration)
#plt.show()
#plt.close()
# run MCMC sampler
print('\n2) Markov-Chain Monte-Carlo sampling\n')
# Define prior
def prior_uniform(x, bounds):
prob = 1 / (bounds[1] - bounds[0])
theta = pso.fit_pso(model,
data,
pars,
states,
bounds,
weights=weights,
maxiter=maxiter,
popsize=popsize,
dist='poisson',
poisson_offset=poisson_offset,
agg=agg,
start_date=start_calibration,
processes=processes)
# theta = np.array([48, 0.01896, 0.02153, 0.02599])
# Assign estimate.
warmup, pars_PSO = assign_PSO(model.parameters, pars, theta)
model.parameters = pars_PSO
# Perform simulation with best-fit results
out = model.sim(end_calibration,
start_date=start_calibration,
warmup=warmup)
# Print statement to stdout once
print(f'\nPSO RESULTS:')
print(f'------------')
print(f'warmup: {warmup}')
print(f'infectivities {pars[1:]}: {theta[1:]}.')
sys.stdout.flush()
# Visualize fit and save in order to check the validity of the first step