Similar sequence matching supporting variable-length and variable-tolerance continuous queries on time-series data stream

Optimizing Cost of Continuous Overlapping Queries over Data Streams by Filter Adaption

“…Conventional continuous query problem in DSMS engages a query sequence that matches the incoming data stream or time series database [16], [17]. Based on that, more complicated query types are defined and discussed, and researchers focus more on optimization of evaluation cost [18].…”

Dynamic time warping constraint learning for large margin nearest neighbor classification

“…DTW distance can overcome this problem by searching an optimal match between two given time series in spite of phase aberration [21]. DTW uses a dynamic programming technique to find the minimal distance between two time series, where sequences are warped by stretching or shrinking the time dimension.…”

Hybrid Lower-Dimensional Transformation for Similar Sequence Matching

“…We performed experiments using four types of data sets: the first one is STOCK-DATA that contains a real stock data [3], [8], [9] consisting of 329,112 entries; the second one is WALK-DATA that contains random walk synthetic data [3], [9] consisting of one million entries; the third one is SINE-DATA that contains synthetic streaming time-series [4] consisting of one million entries; and the fourth one is PERIOD-DATA that contains pseudo periodic synthetic time-series [9] consisting of one million entries. For lower-dimensional transformations, we use two transforms, DFT and PAA, since the characteristics of DCT and Wavelet are similar to those of DFT and Wavelet, respectively [6].…”

“…Time-series data are the sequences of real numbers representing values at specific points in time [1], [3], [5], [7], [8]. The time-series data stored in a database are called data sequences, and those given by users are called query sequences.…”

“…Motivation of the hybrid approach is based on the fact that the well-known transformations concentrate most of the energy into only a few features [1], [2], [8], [10]. Thus, our hybrid approach extracts only a few features from each transformation and integrates the features for a new lower-dimensional transformation.…”