Li Zhang, li.zhang@utrgv.edu, University of Texas Rio Grande Valley
Yifeng Gao, yifeng.gao@utrgv.edu, University of Texas Rio Grande Valley
Yan Zhu, zhuyan@google.com, Google
Jessica Lin, jessica@gmu.edu, George Mason University
Time series patterns have proven to be highly valuable in analyzing large-scale time series in a wide range of applications, due to their ability to extract semantics information, exhibiting a remarkable level of interpretability that aligns seamlessly with human understanding. However, traditional patterns like motifs primarily concentrate on identifying individual patterns, ignoring high-ordered inter-pattern behaviors in time series, which hold significant importance in numerous real-world scenarios. In the last decade, significant research efforts have been dedicated to uncovering high-level semantic structures within time series data for numerous domains and applications. In this tutorial, we review recent research work to develop advanced tools and algorithms for uncovering high-order pattern structures in time series data, enabling the capture of complex semantic relationships. The tutorial is divided into three sections. To begin with, we will explore methodologies tailored for summarizing time series data into concise high-level representations. Following that, we will talk about high-order frequently co-occurring patterns. Lastly, we will discuss evolving patterns that capture systemically evolving behaviors. We will also showcase how to use the discovered patterns in practical applications to guide decision-making in real-world applications.
coming soon...