Tutorial 6: Future regressors

To model future regressors, both past and future values of these regressors have to be known. So in contrast to the lagged regressors in the previous tutorial, future regressors also have a forecasted value for the future in addition to the historic values.

import pandas as pd

# Load the dataset for tutorial 4 with the extra temperature column
df = pd.read_csv("https://github.com/ourownstory/neuralprophet-data/raw/main/kaggle-energy/datasets/tutorial04.csv")
df.head()

	ds	y	temperature
0	2015-01-01	64.92	277.00
1	2015-01-02	58.46	277.95
2	2015-01-03	63.35	278.83
3	2015-01-04	50.54	279.64
4	2015-01-05	64.89	279.05

fig = df.plot(x="ds", y=["y", "temperature"], figsize=(10, 6))

from neuralprophet import NeuralProphet, set_log_level

# Disable logging messages unless there is an error
set_log_level("ERROR")

# Model and prediction
m = NeuralProphet(
    n_changepoints=10,
    yearly_seasonality=True,
    weekly_seasonality=True,
    daily_seasonality=True,
    n_lags=10,
)
m.set_plotting_backend("plotly-static")

# Add the new future regressor
m.add_future_regressor("temperature")

# Continue training the model and making a prediction
metrics = m.fit(df)
forecast = m.predict(df)
m.plot(forecast)

m.plot_components(forecast, components=["future_regressors"])

m.plot_parameters(components=["future_regressors"])

metrics

	MAE	RMSE	Loss	RegLoss	epoch
0	67.432861	79.714500	0.789800	0.0	0
1	40.313995	49.790257	0.434656	0.0	1
2	23.730885	29.712715	0.221153	0.0	2
3	13.360303	17.117233	0.094893	0.0	3
4	10.110947	13.336704	0.061117	0.0	4
...	...	...	...	...	...
95	4.837646	6.482005	0.016463	0.0	95
96	4.830614	6.461479	0.016331	0.0	96
97	4.829221	6.457165	0.016324	0.0	97
98	4.823629	6.465248	0.016352	0.0	98
99	4.838150	6.507308	0.016487	0.0	99

100 rows × 5 columns

df_residuals = pd.DataFrame({"ds": df["ds"], "residuals": df["y"] - forecast["yhat1"]})
fig = df_residuals.plot(x="ds", y="residuals", figsize=(10, 6))