coronavirus-death-rate-recovery-hyperparameter-tuning-202003

results_visualization.py

# pick a set of parameters that resulted in a low loss.
ttd_scale = 63.696418
ttd_shape = 0.968338
ttr_scale = 15.034069
ttr_shape = 1.851922
death_rate = 0.218912

# run the simulation with the parameters.
df_res, loss = run_sim(ttd_scale=ttd_scale, ttd_shape=ttd_shape, ttr_scale=ttr_scale, ttr_shape=ttr_shape, death_rate=death_rate)
loss

# visualize how well it fits recoveries.
sns.lineplot(x="Date", y="New_Recovered", data=df_res, label='New_Recovered')
sns.lineplot(x="Date", y="pred_new_recoveries", data=df_res, label='pred_new_recoveries')

import scipy.stats as stats

# look at the shape of the time to recovery distribution.
x = np.linspace (0, 100, 200)
y = stats.gamma.pdf(x, a=ttr_shape, scale=ttr_scale) 

sns.lineplot(x=x, y=y)

# visualize how well it fits deaths.
sns.lineplot(x="Date", y="New_Deaths", data=df_res, label='New_Deaths')
sns.lineplot(x="Date", y="pred_new_deaths", data=df_res, label='pred_new_deaths')

import scipy.stats as stats

# look at the shape of the time to recovery distribution.
x = np.linspace (0, 100, 200)
y = stats.gamma.pdf(x, a=ttd_shape, scale=ttd_scale) 

sns.lineplot(x=x, y=y)