This book is dedicated to the exploration and explanation of event detection in time series databases. The focus is on events, which are ubiquitous in time series applications where significant behavioral changes are observed at specific points or over time intervals. Event detection is a fundamental function in surveillance and monitoring systems and has been extensively studied over the years. However, this book offers a unified view of the main types of events in time series that every researcher should know: anomalies, change points, and motifs.
The work begins with basic concepts on time series and presents a general taxonomy for event detection. This taxonomy includes: (i) event granularity (point, contextual, and collective), (ii) general strategies (regression, classification, clustering, and model-based methods), (iii) methodological approaches (theory-based and data-driven), (iv) types of machine learning (supervised, semi-supervised, and unsupervised), and (v) data management (the ETL process).
This taxonomy is explored throughout the chapters dedicated to specific event types: anomaly detection, change point detection, and motif discovery. The book also discusses the most advanced evaluation metrics for event detection methods and includes a chapter dedicated to online event detection, addressing the main challenges and strategies (static versus dynamic), including incremental and adaptive learning.
This book is intended for undergraduate and graduate students from different fields who have introductory knowledge of data science or data analysis.
Authors: Eduardo Ogasawara, Rebecca Salles, Fabio Porto, Esther Pacitti
Available at Springer: https://doi.org/10.1007/978-3-031-75941-3