Code

# Separate Train set and Test set
Ntest = 30
train = df.iloc[:-Ntest]
test = df.iloc[-Ntest:]

# Use auto_arima model to find hyperparameter
model = pm.auto_arima(
  train,
  error_action='ignore', 
  trace=True,
  suppress_warnings=True, 
  maxiter=10,
  seasonal=False
)

# Plot predit result function
def plot_result(model, fulldata, train, test):
  params = model.get_params()
  d = params['order'][1]
  train_pred = model.predict_in_sample(start=d, end=-1)
  test_pred, confint = model.predict(n_periods=Ntest, return_conf_int=True)
  fig, ax = plt.subplots(figsize=(10, 5))
  ax.plot(fulldata.index, fulldata, label='data')
  ax.plot(train.index[d:], train_pred, label='fitted')
  ax.plot(test.index, test_pred, label='forecast')
  ax.fill_between(test.index, confint[:,0], confint[:,1], color='red', alpha=0.3)
  ax.legend();

# Plot test result
def plot_test(model, test):
  test_pred, confint = model.predict(n_periods=Ntest, return_conf_int=True)
  fig, ax = plt.subplots(figsize=(10, 5))
  ax.plot(test.index, test, label='true')
  ax.plot(test.index, test_pred, label='forecast')
  ax.fill_between(test.index, confint[:,0], confint[:,1], color='red', alpha=0.3)
  ax.legend();

# RMSE function
def rmse(y, t):
  return np.sqrt(np.mean((t - y)**2))

# ---- output ----

# output model summary
model.summary()

# Plot Result
plot_result(model, df, train, test)

# Plot Test
plot_test(model, test)

# Print RMSE
print("RMSE ARIMA:", rmse(model.predict(Ntest), test))
print("RMSE Naive:", rmse(train.iloc[-1], test))

GOOG

AAPL

IBM

SBUX