DeepAR is a probabilistic forecasting model based on autoregressive recurrent networks. It is designed for large-scale time series forecasting and can handle multiple related time series with covariates.

Key Features

• Probabilistic forecasts (predicts full distribution, not just point estimates)
• Handles multiple time series and covariates
• Scalable to large datasets

Example Use

# Python (GluonTS)
from gluonts.model.deepar import DeepAREstimator
from gluonts.mx.trainer import Trainer

estimator = DeepAREstimator(
    prediction_length=24,
    freq="H",
    trainer=Trainer(epochs=10)
)
      

References

• DeepAR: Probabilistic Forecasting with Autoregressive Recurrent Networks (arXiv)
• GluonTS Documentation