hidden_size=25,
n_rnn_layers=1,
dropout=0.2,
batch_size=16,
n_epochs=20,
optimizer_kwargs={'lr': 1e-3},
model_name='Forecast_LSTM_next_hour',
log_tensorboard=True,
random_state=42
)
# Either train a new model or load best model from checkpoint
if train_new_model==True:
my_model.fit(train_transformed, val_series=val_transformed, verbose=True)
else:
my_model = RNNModel.load_from_checkpoint(model_name='Forecast_LSTM_next_hour', best=True)
# Evaluate Predictions
def eval_model(model):
pred_series = model.predict(n=len(val_transformed))
plt.figure(figsize=(8,5))
val_transformed.plot(label='actual')
pred_series.plot(label='forecast')
plt.title('MAPE: {:.2f}%'.format(mape(pred_series, val_transformed)))
plt.legend()
plt.show()
def backtest(model):
backtest_series = model.historical_forecasts(series_transformed,
start=pd.Timestamp("2021-03-01 00:00:00+00:00"),
my_model.fit(train_transformed, val_series=val_transformed, verbose=True)
def eval_model(model):
pred_series = model.predict(n=96)
plt.figure(figsize=(8, 5))
series_transformed.plot(label='actual')
pred_series.plot(label='forecast')
plt.title('MAPE: {:.2f}%'.format(mape(pred_series, val_transformed)))
plt.legend()
eval_model(my_model)
best_model = RNNModel.load_from_checkpoint(model_name='Eq_RNN', best=True)
eval_model(best_model)
backtest_series = my_model.historical_forecasts(series_transformed,
start=pd.Timestamp('20021231'),
forecast_horizon=12,
retrain=False,
verbose=True)
plt.figure(figsize=(8, 5))
series_transformed.plot(label='actual')
backtest_series.plot(label='backtest')
plt.legend()
plt.title('Backtest, starting Jan 2003, 12-months horizon')
print('MAPE: {:.2f}%'.format(
mape(transformer.inverse_transform(series_transformed),