</>
pythonadvanced

NeuralProphet Deep Time Series Forecast

Forecast complex time series with NeuralProphet combining neural networks and classical decomposition.

python
from neuralprophet import NeuralProphet
import pandas as pd
import numpy as np

dates = pd.date_range('2023-01-01', periods=730, freq='D')
values = 100 + np.cumsum(np.random.randn(730) * 2) + 20 * np.sin(np.linspace(0, 4*np.pi, 730))

df = pd.DataFrame({'ds': dates, 'y': values})

model = NeuralProphet(
    n_forecasts=30,
    n_lags=60,
    yearly_seasonality=True,
    weekly_seasonality=True,
    learning_rate=0.01,
)

df_train, df_val = model.split_df(df, freq='D', valid_p=0.1)
metrics = model.fit(df_train, validation_df=df_val, freq='D')

forecast = model.predict(df)
print(forecast[['ds','yhat1']].tail(5).to_string(index=False))

Use Cases

  • neural forecasting
  • complex trends
  • autoregressive models

Tags

#neuralprophet#forecasting#deep-learning#time-series

Related Snippets

Similar patterns you can reuse in the same workflow.

pythonintermediate

Prophet Time Series Forecasting

Forecast time-series data with Facebook Prophet handling holidays, trends, and seasonality.

Best for: sales forecasting

#prophet#forecasting
pythonintermediate

PyTorch Lightning Training Loop

Simplify PyTorch model training with Lightning's LightningModule for automatic GPU and logging.

Best for: deep learning training

#pytorch-lightning#training
sqlbeginner

Generate Series Calendar Table

Create a date calendar using generate_series for gap-free time series reporting and joins.

Best for: Daily revenue reports

#generate-series#calendar
sqlbeginner

Generate a Date Series

Create a continuous date range to fill gaps in time-series queries.

Best for: Filling gaps in time-series data

#sql#date-series