The field of time series analysis and anomaly detection is rapidly evolving, with a focus on developing innovative and interpretable methods for real-world applications. Recent research has emphasized the importance of capturing complex temporal relationships, handling high-dimensional data, and providing explainable results. Notably, the integration of causal modeling, graph theory, and deep learning techniques has led to significant improvements in anomaly detection and time series forecasting. Furthermore, the development of novel evaluation metrics and frameworks has enabled more accurate assessments of model performance. Overall, the field is moving towards more robust, reliable, and transparent methods for time series analysis and anomaly detection. Noteworthy papers include: Hybrid Autoencoder-Based Framework for Early Fault Detection in Wind Turbines, which introduces a novel ensemble-based deep learning framework for unsupervised anomaly detection. Causal Time Series Modeling of Supraglacial Lake Evolution in Greenland under Distribution Shift, which proposes a regionally-informed causal time-series classification framework for predicting lake evolution.
Advances in Time Series Analysis and Anomaly Detection
Sources
Hybrid Autoencoder-Based Framework for Early Fault Detection in Wind Turbines
Causal Time Series Modeling of Supraglacial Lake Evolution in Greenland under Distribution Shift
STAR: Boosting Time Series Foundation Models for Anomaly Detection through State-aware Adapter
Structured Temporal Causality for Interpretable Multivariate Time Series Anomaly Detection
Diagnosis of Fuel Cell Health Status with Deep Sparse Auto-Encoder Neural Network
Explainable Heterogeneous Anomaly Detection in Financial Networks via Adaptive Expert Routing
Towards Interpretable and Trustworthy Time Series Reasoning: A BlueSky Vision
Modelling complexity in system safety: generalizing the D2T2 methodology
Formally Exploring Time-Series Anomaly Detection Evaluation Metrics
An Encode-then-Decompose Approach to Unsupervised Time Series Anomaly Detection on Contaminated Training Data--Extended Version
Extending Resource Constrained Project Scheduling to Mega-Projects with Model-Based Systems Engineering & Hetero-functional Graph Theory
LMFD: Latent Monotonic Feature Discovery
A Multi-Layer Machine Learning and Econometric Pipeline for Forecasting Market Risk: Evidence from Cryptoasset Liquidity Spillovers
SPAN: Continuous Modeling of Suspicion Progression for Temporal Intention Localization