TY - GEN
T1 - SCFT
T2 - 22nd International Conference on Intelligent Computing, ICIC 2026
AU - Wang, Tianshi
AU - Zhang, Di
N1 - Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2027.
PY - 2027
Y1 - 2027
N2 - Long-horizon multivariate time series forecasting demands a balance between modeling global frequency structures and capturing local irregular fluctuations. We present SCFT, a forecasting framework centered on a novel Dual-Domain Orthogonal Representation that simultaneously exposes global periodicity and preserves temporal shape fidelity. SCFT integrates two supporting mechanisms: (1) an Adaptive Polynomial Approximator that models complex spectral states within the frequency domain, and (2) a Phase-Preserving Local Enhancer that refines short-term temporal irregularities without distorting global frequency alignment. Orthogonality constraints between domains ensure stable and interpretable subspace learning. Extensive evaluations across multiple benchmarks demonstrate that SCFT achieves competitive predictive accuracy while maintaining superior trend consistency, parameter efficiency, and predictable memory scaling compared to heavy baselines.
AB - Long-horizon multivariate time series forecasting demands a balance between modeling global frequency structures and capturing local irregular fluctuations. We present SCFT, a forecasting framework centered on a novel Dual-Domain Orthogonal Representation that simultaneously exposes global periodicity and preserves temporal shape fidelity. SCFT integrates two supporting mechanisms: (1) an Adaptive Polynomial Approximator that models complex spectral states within the frequency domain, and (2) a Phase-Preserving Local Enhancer that refines short-term temporal irregularities without distorting global frequency alignment. Orthogonality constraints between domains ensure stable and interpretable subspace learning. Extensive evaluations across multiple benchmarks demonstrate that SCFT achieves competitive predictive accuracy while maintaining superior trend consistency, parameter efficiency, and predictable memory scaling compared to heavy baselines.
KW - Kolmogorov-Arnold networks
KW - Lightweight architecture
KW - Spectral decomposition
KW - Time series forecasting
UR - https://www.scopus.com/pages/publications/105046843660
U2 - 10.1007/978-981-92-3551-3_22
DO - 10.1007/978-981-92-3551-3_22
M3 - Conference Proceeding
AN - SCOPUS:105046843660
SN - 9789819235506
T3 - Communications in Computer and Information Science
SP - 261
EP - 273
BT - Advanced Intelligent Computing Technology and Applications - 22nd International Conference on Intelligent Computing, ICIC 2026, Proceedings
A2 - Huang, De-Shuang
A2 - Zhang, Chuanlei
A2 - Chen, Wei
A2 - Li, Bo
A2 - Zhang, Qinhu
A2 - Bao, Wenzheng
A2 - Pan, Yijie
A2 - Premaratne, Prashan
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 22 July 2026 through 26 July 2026
ER -