The field of time series forecasting is rapidly advancing with the integration of innovative techniques and models. Recent developments have focused on improving the accuracy and robustness of forecasting models, particularly in handling complex temporal dependencies and non-stationary data. The incorporation of quantum-optimized approaches, multimodal alignment, and adaptive graph learning has shown promising results in enhancing forecasting performance. Additionally, the use of large language models and ensemble methods has improved the ability to capture long-range dependencies and provide uncertainty-aware predictions. Noteworthy papers include Quantum-Optimized Selective State Space Model, BALM-TSF, and ST-Hyper, which have demonstrated state-of-the-art performance in various benchmark datasets. These advancements have significant implications for real-world applications, such as logistical demand-supply forecasting, parking availability prediction, and wind power forecasting.
Advancements in Time Series Forecasting
Sources
Quantum-Optimized Selective State Space Model for Efficient Time Series Prediction
BALM-TSF: Balanced Multimodal Alignment for LLM-Based Time Series Forecasting
Robust Spatiotemporal Forecasting Using Adaptive Deep-Unfolded Variational Mode Decomposition
Text Reinforcement for Multimodal Time Series Forecasting
TimeCopilot
Forecasting the Ionosphere from Sparse GNSS Data with Temporal-Fusion Transformers
Towards Trustworthy Vital Sign Forecasting: Leveraging Uncertainty for Prediction Intervals
StoxLSTM: A Stochastic Extended Long Short-Term Memory Network for Time Series Forecasting
When LLM Meets Time Series: Can LLMs Perform Multi-Step Time Series Reasoning and Inference
ACA-Net: Future Graph Learning for Logistical Demand-Supply Forecasting
RDIT: Residual-based Diffusion Implicit Models for Probabilistic Time Series Forecasting
ST-Hyper: Learning High-Order Dependencies Across Multiple Spatial-Temporal Scales for Multivariate Time Series Forecasting
Reservoir Predictive Path Integral Control for Unknown Nonlinear Dynamics
MillGNN: Learning Multi-Scale Lead-Lag Dependencies for Multi-Variate Time Series Forecasting
One-Embedding-Fits-All: Efficient Zero-Shot Time Series Forecasting by a Model Zoo
Rethinking the long-range dependency in Mamba/SSM and transformer models
Parking Availability Prediction via Fusing Multi-Source Data with A Self-Supervised Learning Enhanced Spatio-Temporal Inverted Transformer
Echo State Networks as State-Space Models: A Systems Perspective
Echoes Before Collapse: Deep Learning Detection of Flickering in Complex Systems
VARMA-Enhanced Transformer for Time Series Forecasting
BEDTime: A Unified Benchmark for Automatically Describing Time Series
Select, then Balance: A Plug-and-Play Framework for Exogenous-Aware Spatio-Temporal Forecasting
ARIES: Relation Assessment and Model Recommendation for Deep Time Series Forecasting
WindFM: An Open-Source Foundation Model for Zero-Shot Wind Power Forecasting
Adaptive Rainfall Forecasting from Multiple Geographical Models Using Matrix Profile and Ensemble Learning
Improving Greenland Bed Topography Mapping with Uncertainty-Aware Graph Learning on Sparse Radar Data
MAESTRO: Multi-modal Adaptive Ensemble for Spectro-Temporal Robust Optimization