Learning DTW Global Constraint for Time Series Classification - Computer Science > Artificial IntelligenceReport as inadecuate




Learning DTW Global Constraint for Time Series Classification - Computer Science > Artificial Intelligence - Download this document for free, or read online. Document in PDF available to download.

Abstract: 1-Nearest Neighbor with the Dynamic Time Warping DTW distance is one of themost effective classifiers on time series domain. Since the global constrainthas been introduced in speech community, many global constraint models havebeen proposed including Sakoe-Chiba S-C band, Itakura Parallelogram, andRatanamahatana-Keogh R-K band. The R-K band is a general global constraintmodel that can represent any global constraints with arbitrary shape and sizeeffectively. However, we need a good learning algorithm to discover the mostsuitable set of R-K bands, and the current R-K band learning algorithm stillsuffers from an -overfitting- phenomenon. In this paper, we propose two newlearning algorithms, i.e., band boundary extraction algorithm and iterativelearning algorithm. The band boundary extraction is calculated from the boundof all possible warping paths in each class, and the iterative learning isadjusted from the original R-K band learning. We also use a Silhouette index, awell-known clustering validation technique, as a heuristic function, and thelower bound function, LB Keogh, to enhance the prediction speed. Twentydatasets, from the Workshop and Challenge on Time Series Classification, heldin conjunction of the SIGKDD 2007, are used to evaluate our approach.



Author: Vit Niennattrakul, Chotirat Ann Ratanamahatana

Source: https://arxiv.org/



DOWNLOAD PDF




Related documents