The field of time series forecasting is witnessing a significant shift towards the integration of deep learning and traditional signal processing techniques. Researchers are exploring new avenues to combine the strengths of both worlds, resulting in more accurate and robust forecasting models. A key area of focus is the development of novel architectures that can effectively capture temporal dependencies, spatial relationships, and multi-scale periodicity in time series data. Notably, the use of Fourier and Laplace transforms, as well as large language models, is becoming increasingly popular in this context. Furthermore, researchers are also investigating the application of Vision Foundation Models for spatio-temporal forecasting, which has shown promising results. Overall, these developments are advancing the field of time series forecasting and opening up new possibilities for real-world applications. Noteworthy papers include: Fourier Basis Mapping, which proposes a novel time-frequency learning framework for time series forecasting, and Reprogramming Vision Foundation Models for Spatio-Temporal Forecasting, which introduces a framework for adapting Vision Foundation Models to spatio-temporal forecasting tasks.
Emerging Trends in Time Series Forecasting
Sources
Fourier Basis Mapping: A Time-Frequency Learning Framework for Time Series Forecasting
EV-STLLM: Electric vehicle charging forecasting based on spatio-temporal large language models with multi-frequency and multi-scale information fusion
Frequency-aware Surrogate Modeling With SMT Kernels For Advanced Data Forecasting
Fusing Large Language Models with Temporal Transformers for Time Series Forecasting
TAT: Temporal-Aligned Transformer for Multi-Horizon Peak Demand Forecasting
LLMs Meet Cross-Modal Time Series Analytics: Overview and Directions
A Generalizable Physics-Enhanced State Space Model for Long-Term Dynamics Forecasting in Complex Environments
Data Augmentation in Time Series Forecasting through Inverted Framework
Reprogramming Vision Foundation Models for Spatio-Temporal Forecasting
Targeted Mining of Time-Interval Related Patterns
FLDmamba: Integrating Fourier and Laplace Transform Decomposition with Mamba for Enhanced Time Series Prediction
The Power of Architecture: Deep Dive into Transformer Architectures for Long-Term Time Series Forecasting