Querying and mining of time series data

Abstract

The last decade has witnessed a tremendous growths of interests in applications that deal with querying and mining of time series data. Numerous representation methods for dimensionality reduction and similarity measures geared towards time series have been introduced. Each individual work introducing a particular method has made specific claims and, aside from the occasional theoretical justifications, provided quantitative experimental observations. However, for the most part, the comparative aspects of these experiments were too narrowly focused on demonstrating the benefits of the proposed methods over some of the previously introduced ones. In order to provide a comprehensive validation, we conducted an extensive set of time series experiments re-implementing 8 different representation methods and 9 similarity measures and their variants, and testing their effectiveness on 38 time series data sets from a wide variety of application domains. In this paper, we give an overview of these different techniques and present our comparative experimental findings regarding their effectiveness. Our experiments have provided both a unified validation of some of the existing achievements, and in some cases, suggested that certain claims in the literature may be unduly optimistic.

Oscillatory patterns in the light curves of five long-term monitored type 1 active galactic nuclei

“…The time series investigation methods can be classified as shape-based, structure-based (or model-based), and dimensionality reduction (Ding et al 2008). The dimensionality reduction methods are based on data transformation such as discrete Fourier transformation, single value decomposition, continuous wavelet transformation (CWT) and discrete wavelet transformation (DWT), piecewise approximation, and Chebyshev polynomials.…”

DTWDIR: An Enhanced DTW Algorithm for Autistic Child Behaviour Monitoring

“…The interpolation pre-processing may change the original data values (raw data) but it keeps the form of the signal pattern [13] therefore DTW accuracy in interpolation is less than in raw data attributes as in Table 3 because the purpose of DTW algorithm matches the best wrapping path between the two signals [14] but the DTWDir accuracy was not affected. DTW accuracy in 5 sec is better than in interpolation but it still was not higher than raw data pre-processing.…”

Pattern Recognition in Time Series

“…A common variation of this algorithm predicts the classification of the test object to be the most common classification found among the "k" nearest neighbors in the training set (k-NN). The 1-NN classifier, with leaving-one-out cross validation, has become the standard method used to compare and evaluate the utilities of time series representations and similarity measures [Ding08].…”

“…From this experiment, we can conclude that SAX is competitive with Euclidean distance, but requires far less space. More experiments that compare different time series representations can be found in [Ding08]. …”