N-BEATSx
N-BEATSx: Neural Basis Expansion Analysis for Interpretable Time Series Forecasting · Also known as: N-BEATSx, NBEATS-x
N-BEATSx is an extension of the N-BEATS neural time series forecasting model that incorporates exogenous (external) variables through a cross-learner architecture. Published in 2023, N-BEATSx improves upon N-BEATS by enabling the model to leverage additional features beyond the historical time series values.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
N-BEATSx is ideal for forecasting when relevant exogenous features are available and can improve predictions. Use it for energy, traffic, and financial forecasting where external variables matter. If only historical series data is available, the original N-BEATS is sufficient. Use statistical methods like ARIMA when data is limited and interpretability is paramount.
Strengths & limitations
- Naturally incorporates exogenous variables through a theoretically motivated architecture
- Maintains interpretability by explicitly modeling trend and seasonality components
- Handles variable-length sequences and irregular sampling through neural basis expansion
- Achieves strong performance on diverse forecasting benchmarks with external features
- Requires sufficient data to learn complex trend and seasonal patterns; poor performance on very short series
- Interpretability, while better than black-box models, is still limited compared to classical decomposition methods
- Exogenous feature engineering remains necessary; the model does not automatically discover relevant features
Frequently asked
What are exogenous variables and why include them?
Exogenous variables are external features that influence the target variable but are not derived from the target itself. Examples include weather for energy demand, holidays for retail sales, or economic indicators for financial forecasting. Including exogenous variables typically improves forecast accuracy when they have true predictive power, allowing the model to capture dependencies beyond pure time series patterns.
How does N-BEATSx differ from classical forecasting with external features?
Classical methods like regression assume linear relationships between variables. N-BEATSx learns nonlinear interactions through neural networks, enabling more complex dependencies. The cross-learner architecture enables the model to learn task-specific transformations of exogenous features rather than using them as-is, improving flexibility and predictive power.
What is basis expansion and why use it?
Basis expansion represents complex functions as linear combinations of simpler basis functions. N-BEATSx learns which basis functions (trends, seasonality patterns) are important for the specific forecasting task. This is more efficient than using raw neural network layers and provides interpretability by decomposing forecasts into meaningful components.
Sources
- Challu, C., Olivares, K. Q., Oreshkin, B., Garza, F., Mergenthaler-Canseco, M., & Dubrawski, A. (2023). N-BEATSx: Neural Basis Expansion Analysis for Interpretable Time Series Forecasting. In ICLR 2023 Workshop on Multimodal Learning for Science (p. 4). link ↗
How to cite this page
ScholarGate. (2026, June 3). N-BEATSx: Neural Basis Expansion Analysis for Interpretable Time Series Forecasting. ScholarGate. https://scholargate.app/en/deep-learning/n-beatsx
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Mamba (State Space Model)Deep learning↔ compare
- Spatial-Temporal GCNDeep learning↔ compare
- TimeGPTDeep learning↔ compare
- Vision MambaDeep learning↔ compare