High performance discovery in time series

More about the book

Time-series data, which arrives in a sequential order, is prevalent in various fields such as physics, finance, music, networking, and medical instrumentation. Developing fast, scalable algorithms for analyzing single or multiple time series can facilitate scientific discoveries, medical diagnoses, and potential profits. This work introduces rapid-discovery techniques for identifying segments of time series with numerous events, as well as for detecting closely related time series. Unlike typical methods that focus on predicting future points through regression analysis, the emphasis here is on efficient discovery within time series, showcasing novel algorithmic contributions alongside practical algorithms and case studies. It requires only a basic understanding of calculus and some linear algebra. Key features include algorithms for uncovering unusual activity bursts in large databases, exploring correlation relationships among extensive time series, and adapting techniques for individual needs. Additionally, it covers algorithms for query by humming, gamma-ray burst detection, pairs trading, and density detection, along with self-contained descriptions of wavelets and fast Fourier transforms. This monograph serves as a comprehensive resource for computer scientists, physicists, medical researchers, financial mathematicians, musicologists, and graduate students in data-rich fields.