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

2026-07-16

AI Summary

This study addresses the limitations of existing mixed-frequency time series forecasting methods, which predominantly rely on linear frequency alignment and thus fail to capture nonlinear inter-variable relationships, ultimately constraining predictive accuracy. To overcome this, we propose a depthwise separable neural network architecture that models distinct frequency alignment pathways separately while incorporating a cross-stage parameter sharing mechanism, enabling efficient fusion of high-frequency information and stable training of deep networks. Our work is the first to integrate nonlinear deep learning into mixed-frequency alignment frameworks, establishing corresponding approximation theory and deriving non-asymptotic prediction error bounds that transcend the constraints of traditional linear approaches. Empirical results demonstrate that the proposed method significantly outperforms state-of-the-art models under limited sample sizes, achieving superior performance in forecasting U.S. quarterly macroeconomic indicators.

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.

Problem

• mixed-frequency
• time series forecasting
• frequency alignment
• nonlinearity
• prediction accuracy

Innovation

• mixed-frequency time series
• depth-separable neural network
• frequency alignment
• nonlinear transformation
• parameter sharing