23 May 2025 in IEEE Access
Global warming accelerates permafrost degradation, compromising the reliability of critical infrastructure relied upon by over five million people daily. Additionally, permafrost thaw releases substantial methane emissions due to the thawing of swamps, further amplifying global warming and climate change and thus posing a significant threat to more than eight billion people worldwide. To mitigate this growing risk, policymakers and stakeholders need accurate predictions of permafrost thaw progression. Comprehensive physics-based permafrost models often require complex, location-specific fine-tuning, making them impractical for widespread use. Although simpler models with fewer input parameters offer convenience, they generally lack accuracy. Purely data-driven models also face limitations due to the spatial and temporal sparsity of observational data. This work develops a physics-informed machine learning framework to predict permafrost thaw rates. By integrating a physics-based model into machine learning, the framework significantly enhances the feature set, enabling models to train on higher-quality data. This approach improves permafrost thaw rate predictions, supporting more reliable decision-making for construction and infrastructure maintenance in permafrost-vulnerable regions, with a forecast horizon spanning several decades.
11 September 2024 in IEEE Access
Automated Machine Learning (AutoML) tools for time series forecasting represent a frontier in both academic and industrial research, addressing the need for efficient, accurate predictions in various domains. This study focuses on the development of Automated Time Series Forecasting (AutoTS), specifically in the domains of demand forecasting and traffic prediction. Through a comprehensive empirical evaluation and comparative analysis, this study investigates the performance of 35 time series forecasting methods and scrutinizes the forecast quality and time series characteristics. In the experiments, we used six real and one synthetic datasets with 325 time series in total. The forecasting for four distinct horizons is studied. From our large-scale empirical study, we draw the following main conclusions: boosting-based methods, which are often overlooked, have a strong performance; the global modeling approach is promising because it provides competitive performance with a small computational cost; meta-learning via portfolio selection performs better than one based on meta-features. We hope that our empirical study will pave the way for more efficient AutoML systems for time series forecasting.