Mixed-Frequency Time Series Forecasting via Depth-Separable Neural Networks

Abstract

To better forecast mixed-frequency time series, it is the key to choose a suitable way for frequency alignment. However, the existing methods are all limited to linear transformations, and this may overlook the possible nonlinearity, leading to a worse prediction. We alternatively consider a deep neural network for each frequency alignment, and hence a depth-separable neural network. Moreover, a parameter-sharing mechanism is adopted across the alignment at each stage, making possible a deeper network for a large set of higher-frequency predictors. This paper establishes an approximation theory for the proposed depth-separable network, and a non-asymptotic prediction error bound is also derived. Simulation studies demonstrate the finite-sample performance of the proposed method, and an empirical application to forecasting U.S. quarterly macroeconomic variables using monthly and daily indicators, highlights its superior predictive accuracy over existing mixed-frequency methods.

S²-PINN: Stochastic Separable Physics-Informed Neural Networks
Zhendong Li, Akwum Onwunta
arXiv2 Oct 2026PreprintCC BY 4.0Full textMathematical physicsarXiv
Forecasting seasonal influenza epidemics with physics-informed neural networks
Martina Rama, Gabriele Santin, Giulia Cencetti +2
Epidemics29 May 2026Peer-reviewedCC BY-NC-ND5 citedGeneral mathematicsOpenAlex
Explainable spatiotemporal Graph Neural Networks: Multi-City PM2.5 forecasting in Bangladesh with geographic graph topology validation
Pritom Ray Nobin
PLOS One12 Oct 2026Peer-reviewedCC BY 4.0Full textEnvironmental chemistry Atmospheric science & meteorologyPLOS
Sparse-Observation Atmospheric Thermal Forecasting with Physics-Informed Neural Networks for Climate-Aware Digital Twins
Tannaz Goodarzvand Chegini, Elyas Shivanian, Behzad Karimi +1
arXiv23 Sep 2026PreprintarXiv licenceFull textAnalysis & PDEs Numerical analysis & scientific computingarXiv
Inverse Cross-spectral Neural Networks for Multivariate Time Series
Lorenzo Marinucci, Leonardo Di Nino, Gabriele D'Acunto +2
arXiv5 Oct 2026PreprintarXiv licenceFull textStatisticsarXiv

Cite

BibTeX RIS

BibTeX

@misc{wang2026mixedfrequency,
  title = {{Mixed-Frequency Time Series Forecasting via Depth-Separable Neural Networks}},
  author = {Yize Wang and Qianqian Zhu and Guodong Li},
  howpublished = {arXiv},
  year = {2026},
  month = oct,
  doi = {10.48550/arxiv.2607.14771},
  eprint = {2607.14771},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2607.14771},
}

RIS

TY  - UNPB
TI  - Mixed-Frequency Time Series Forecasting via Depth-Separable Neural Networks
AU  - Yize Wang
AU  - Qianqian Zhu
AU  - Guodong Li
PY  - 2026
DA  - 2026/10/07
PB  - arXiv
DO  - 10.48550/arxiv.2607.14771
UR  - https://arxiv.org/abs/2607.14771
AB  - To better forecast mixed-frequency time series, it is the key to choose a suitable way for frequency alignment. However, the existing methods are all limited to linear transformations, and this may overlook the possible nonlinearity, leading to a worse prediction. We alternatively consider a deep neural network for each frequency alignment, and hence a depth-separable neural network. Moreover, a parameter-sharing mechanism is adopted across the alignment at each stage, making possible a deeper network for a large set of higher-frequency predictors. This paper establishes an approximation theory for the proposed depth-separable network, and a non-asymptotic prediction error bound is also derived. Simulation studies demonstrate the finite-sample performance of the proposed method, and an empirical application to forecasting U.S. quarterly macroeconomic variables using monthly and daily indicators, highlights its superior predictive accuracy over existing mixed-frequency methods.
ER  -

Abstract

Related

Cite