A Survey on Irregular Time Series Analysis: Taxonomies, Methods and Future Directions
Rujiao Wang, Tianzi Zang, Weiwei Yuan, Wenwen Zheng, Tianhui Chen · 2025
Time series analysis involves interpreting discrete, time-indexed observations to understand and predict the dynamics of complex systems, playing a pivotal role in decision-making across various domains. However, unlike idealized regular time series, real-world data frequently present as irregular time series characterized by variable sampling intervals and asynchronous observations across channels. These irregularities pose significant challenges to regular time series modeling approaches, as unpredictable sampling patterns and channel asynchrony disrupt the extraction of temporal patterns and may even contain essential information about the underlying processes. Thus, irregular time series analysis has attracted significant attention in recent years, leading to a surge of innovative methods and frameworks. In this survey, we provide an up-to-date and comprehensive review of irregular time series analysis methods. We first introduce two taxonomies from the perspectives of data representation and modeling methods. The representation taxonomy summarizes and compares existing representations for irregular time series, while the treelike methods taxonomy organizes and guides our detailed discussions and comparative analyses of existing methods. We also compile commonly used datasets and publicly available code resources to support researchers entering this field. Finally, we outline some promising future research directions to inspire further advancements in irregular time series analysis.