A Bag-of-Features Framework to Classify Time Series

M. G. Baydogan, G. Runger, E. Tuv
2013 IEEE Transactions on Pattern Analysis and Machine Intelligence  
Time series classification is an important task with many challenging applications. Nearest-neighbor classifiers with Dynamic Time Warping (DTW) distance is a strong solution in this context, but its performance degrades with long time series, relatively short features of interest, and moderate noise. On the other hand, feature-based ap-* Corresponding author. Email: mbaydoga@asu.edu c 2012 IEEE 1 easily integrated through a fast, efficient learner that handles mixed data types, different
more » ... etc., and relevant global features can easily supplement the codebook in our framework. We compare our classifier to well-known nearest-neighbor classifiers, with dynamic time warping distance measures, and shapelet methods. Our experimental results show that TSBF provides better results than competitive methods on benchmark data sets from the UCR time series database.
doi:10.1109/tpami.2013.72 pmid:24051736 fatcat:pxwoupvonnb2rno545kbwizm34