Fusion Methods for Time-Series Classification

Time-series classification is the common theoretical background of many recognition tasks performed by computers, such as handwriting recognition, speech recognition or detection of abnormalities in electrocardiograph signals. In this book, the state-of-the-art in time-series classification is surveyed and five new techniques are presented. Four out of them aim at making the recognition more accurate, while the proposed instance-selection algorithm speeds up time-series classification. Besides time-series classification tasks, potential applications of the proposed techniques include problems from various domains, e.g. web science or medicine.

Autorentext

Krisztian Antal Buza obtained his diploma at the Budapest University of Technology and Economics in 2007, and his PhD at the University of Hildesheim in 2011. His work on time-series classification was honored by the Best Paper Award at the renowned conference on Computational Science and Engineering of the Institute of Electrical and Electronics Engineers (IEEE) in 2010.

Inhalt

Contents: Survey of the state-of-the-art in time-series classification - Individual Quality Estimation - Speeding-up time series classification using instance selection - GRAMOFON, a graph-based ensemble framework - Fusion of time series distance measures - Discovery of recurrent patterns (motifs) in time series - Applications to electrocardiograph signals and web-science problems.