New AI methods tackle time series forecasting and model explainability · 5 sources tracked
ByPulseAugur Editorial·[11 sources]·
Researchers have introduced KARMA, a novel method for explaining time-series forecasting models by constructing a Markov surrogate model that captures temporal dependencies. This approach identifies the minimal history length required for predictive sufficiency and estimates a Markov transition kernel, offering a five-level global explanation hierarchy. Separately, a framework called The Simulacrum uses decision-theoretic pretraining to develop neural network-based time series estimators that can approximate optimal decision rules and achieve competitive forecasting accuracy on real-world benchmarks. Additionally, a study on catastrophic forgetting in time series foundation models indicates that while fine-tuning improves accuracy, it can lead to forgetting, though larger models show more robustness and mitigation techniques can help smaller models match performance.
AI
IMPACT
These advancements offer improved accuracy and interpretability for time series models, potentially impacting fields like finance, weather prediction, and autonomous driving.
RANK_REASON
Cluster contains multiple research papers on novel AI methods for time series forecasting and explainability.
arXiv:2607.01918v1 Announce Type: new Abstract: We present Zeus, a unified tuning-free Time Series Foundation Model (TSFM) that delivers superior performance across diverse analysis tasks without any task-specific fine-tuning. Unlike prior studies that primarily focus on zero-sho…
We present Zeus, a unified tuning-free Time Series Foundation Model (TSFM) that delivers superior performance across diverse analysis tasks without any task-specific fine-tuning. Unlike prior studies that primarily focus on zero-shot forecasting but require task-specific tuning f…
arXiv cs.LG
TIER_1English(EN)·Haroon Gharwi, Yue Dai, Kai Shu·
arXiv:2607.00197v1 Announce Type: new Abstract: Long-horizon multivariate time series forecasting (LTSF) remains challenging due to non-stationarity, regime shifts, and error accumulation. The Variability-Aware Recursive Neural Network (VARNN) is designed to track such variabilit…
arXiv:2606.28670v1 Announce Type: cross Abstract: We introduce MACROCAST, a lightweight Time Series Foundation Model (TSFM) for real-time macroeconomic forecasting. Existing TSFMs suffer from data leakage in two forms: temporal contamination, as the model may have seen the realiz…
arXiv cs.LG
TIER_1English(EN)·Oleksandr Shchur, Abdul Fatir Ansari, Caner Turkmen, Lorenzo Stella, Nick Erickson, Pablo Guerron, Michael Bohlke-Schneider, Yuyang Wang·
arXiv:2509.26468v3 Announce Type: replace Abstract: Benchmark quality is critical for meaningful evaluation and sustained progress in time series forecasting, particularly with the rise of pretrained models. Existing benchmarks often have limited domain coverage or overlook real-…
arXiv:2606.27711v1 Announce Type: cross Abstract: We introduce a neural network-based framework for learning time series estimators through a process we term decision-theoretic pretraining. Analysts specify a generative world, a distribution over data-generating processes, and a …
arXiv:2510.00809v3 Announce Type: replace Abstract: While Time Series Foundation Models (TSFMs) excel in zero-shot tasks, their behavior under continual fine tuning is poorly understood. We present the first systematic study of catastrophic forgetting in TSFMs (TimesFM-2.0, Chron…
arXiv:2606.27599v1 Announce Type: cross Abstract: While many explainable AI (XAI) methods have been proposed, most are not designed for time-series forecasting models and often rely on the implicit assumption that timestamp features are independent. This assumption ignores the fu…
We introduce a neural network-based framework for learning time series estimators through a process we term decision-theoretic pretraining. Analysts specify a generative world, a distribution over data-generating processes, and a target decision objective. A neural network traine…
arXiv cs.LG
TIER_1English(EN)·Qiyuan Wu, Katie Z Luo, Bharath Hariharan, Wei-Lun Chao, Mark Campbell·
arXiv:2606.26424v1 Announce Type: new Abstract: Trajectory forecasting for autonomous driving has advanced rapidly, yet representative models often produce uninformative posteriors over forecast modes, causing problems for mode pruning. We trace this to a modeling-training mismat…
arXiv:2412.19897v3 Announce Type: replace Abstract: We introduce a local surrogate approach for explainable time-series forecasting. An initially non-interpretable predictive model to improve the forecast of a classical time-series 'base model' is used. 'Explainability' of the co…
