Neuralized multi-wavelet method improves time series classification and forecasting
Neuralized Multi-Wavelet Decomposition for Time Series Classification and Forecasting
Machine LearningArtificial Intelligence
Summary
Time series data, like stock prices or weather patterns, often contain complex patterns at different time scales and frequencies. The authors identify a problem where existing methods look at time patterns or frequency patterns separately, which limits how well they understand the data. They designed a new deep learning approach called m-WCN that combines these two perspectives, learning patterns across multiple time and frequency scales jointly. Their approach led to clearer, interpretable features and improved accuracy in identifying and predicting time series behaviors on many datasets.
What this means in practice
- •For financial data teams: Improve forecasting of stock or economic indicators by jointly capturing temporal and frequency patterns using neuralized multi-wavelet representations.
- •For healthcare analytics teams: Enhance diagnosis support by classifying complex physiological time series with frequency-aware, multi-resolution features.
Authors
Xiaohan Jiang, Jingyuan Wang, Jiahao Ji, Yongyao Wang, Chen Yang, Junjie Wu
Abstract
Time series analysis is fundamental in domains such as finance, healthcare, and meteorology. Real-world time series often exhibit multiscale characteristics shaped by diverse latent factors, resulting in intricate temporal patterns and rich frequency structures. However, existing approaches typically focus on either frequency-domain decomposition or time-domain pattern extraction in isolation, neglecting their joint structure. This decoupled modeling limits representation expressiveness and undermines performance in tasks requiring simultaneous temporal and spectral reasoning. To address this gap, we propose m-WCN, a novel end-to-end deep learning framework that neuralizes multi-wavelet decomposition for joint extraction of temporal patterns and frequency components. By approximating the classical GHM multi-wavelet transform with trainable convolutional operators and enforcing orthogonality constraints, m-WCN produces interpretable multi-resolution representations. Built on this foundation, we introduce two task-specific architectures: TFBC for time series classification, which boosts discriminative features across frequency scales, and FTB for forecasting, which ensembles frequency-aware predictors. Extensive experiments on 64 UCR datasets and seven public forecasting benchmarks demonstrate the effectiveness of our approach. Built on the neuralized m-WCN, our TFBC and FTB outperform various baseline models across diverse datasets, achieving average improvements of 19.97% in classification and 19.92% in forecasting tasks.