The field of time series forecasting is rapidly advancing with a focus on improving accuracy and efficiency. Recent developments have seen the integration of various techniques such as multi-step forecasting, graph neural networks, and attention mechanisms to better capture complex patterns and relationships in time series data. Notably, the use of hybrid models combining statistical and machine learning approaches has shown promise in handling both linear and nonlinear patterns. Furthermore, innovations in areas like dynamic patch encoding, entropy-guided patching, and non-autoregressive forecasting are pushing the boundaries of what is possible in time series prediction. Some particularly noteworthy papers include the proposal of Echo Flow Networks, which achieve state-of-the-art performance while reducing training time and model size, and the introduction of TimeEmb, a lightweight framework for static-dynamic disentanglement that outperforms existing baselines while requiring fewer computational resources.
Advances in Time Series Forecasting
Sources
Improving accuracy in short mortality rate series: Exploring Multi-step Forecasting Approaches in Hybrid Systems
Multi-Scale Spatial-Temporal Hypergraph Network with Lead-Lag Structures for Stock Time Series Forecasting
Graph Neural Networks with Diversity-aware Neighbor Selection and Dynamic Multi-scale Fusion for Multivariate Time Series Forecasting
IndexNet: Timestamp and Variable-Aware Modeling for Time Series Forecasting
Echo Flow Networks
Neuroplasticity-inspired dynamic ANNs for multi-task demand forecasting
DSAT-HD: Dual-Stream Adaptive Transformer with Hybrid Decomposition for Multivariate Time Series Forecasting
STCast: Adaptive Boundary Alignment for Global and Regional Weather Forecasting
Integrated Forecasting of Marine Renewable Power: An Adaptively Bayesian-Optimized MVMD-LSTM Framework for Wind-Solar-Wave Energy
WDformer: A Wavelet-based Differential Transformer Model for Time Series Forecasting
Joint Embeddings Go Temporal
Swift: An Autoregressive Consistency Model for Efficient Weather Forecasting
ReNF: Rethinking the Design Space of Neural Long-Term Time Series Forecasters
EntroPE: Entropy-Guided Dynamic Patch Encoder for Time Series Forecasting
TimeEmb: A Lightweight Static-Dynamic Disentanglement Framework for Time Series Forecasting
A Neuro-Fuzzy System for Interpretable Long-Term Stock Market Forecasting
KAIROS: Unified Training for Universal Non-Autoregressive Time Series Forecasting
Efficiently Generating Correlated Sample Paths from Multi-step Time Series Foundation Models