# Start Testing
# Create an instance of the filter
kf = Kalman(history, order)
# Use the first 7 of them for training
obs_train = obs[:history]
model_train = model[:history]
# The rest to be used dynamically
obs_dyn = obs[history:]
model_dyn = model[history:]
fcst = np.zeros_like(obs_dyn)
# Perform an initial training of the model
kf.train_me(obs_train, model_train)
for ij in range(len(obs_dyn)):
# Provide a correction to the forecast
fcst[ij] = kf.adjust_forecast(model_dyn[ij])
# Update filter
kf.train_me([obs_dyn[ij]], [model_dyn[ij]])
fig = plt.figure(figsize=figSize)
plt.plot(obs_dyn[pltRange], label='obs')
plt.plot(model_dyn[pltRange], label='model')
plt.plot(fcst[pltRange], label='kalman')
plt.legend()
plt.show()
