Experiencing SAX: a novel symbolic representation of time series

By Jessica Lin, Eamonn J. Keogh, Li Wei, Stefano Lonardi. Data mining and knowledge discovery, 2007

1,732 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Economics and Econometrics, Signal Processing

View Paper: https://www.semanticscholar.org/paper/244e27bf8c5a5277a733c5271fdeb306f98115d5

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Economics and Econometrics

In Signal Processing

Citations in the corpus

The great time series classification bake off: a review and experimental evaluation of recent algorithmic advances (2016, methodology): “Instead of a full enumerative search at each node, the fast shapelets algorithm discretises and approximates the shapelets using a symbolic aggregate approximation (SAX) (Lin et al. 2007).”
Keeping it Short and Simple: Summarising Complex Event Sequences with Multivariate Patterns (2015, methodology): “We then discretised each z-normalised attribute into 5 intervals using SAX [13].”
Outlier Detection for Temporal Data: A Survey (2014, methodology): “To make the computations efﬁcient, the TARZAN algorithm was proposed which exploits suf-ﬁx trees [68], [70], [71].”
Clustering of Electricity Consumption Behavior Dynamics Toward Big Data Applications (2016, background): “Generally, the time domain is partitioned uniformly, and the amplitude axis is partitioned based on the normal distribution hypothesis [25], [40].”
Querying and mining of time series data: experimental comparison of representations and distance measures (2008, background): “In contrast, in Figure 5 we can see that highly periodic data can slightly favor the spectral representations (DCT, DFT, CHEB) over the polynomial representations (SAX, APCA, DWT/PAA, IPLA).”
iSAX: indexing and mining terabyte sized time series (2008, methodology): “• Non-Lower Bounding [1][7][13] • SAX* [12], iSAX* o Trees • Non-Data Adaptive o Wavelets* o Random Mappings o Spectral DFT* [6] DCT* Chebyshev Polynomials* [4] o Piecewise Aggregate Approximation* [9] • Data Dictated o Clipped Data*”
LSTM Fully Convolutional Networks for Time Series Classification (2017, background): “Bag-of-Words (BoW) [2], Bag-of-features (TSBF) [3], Bag-of-SFA-Symbols (BOSS) [4], BOSSVS [5], Word ExtrAction for time Series cLassiﬁcation (WEASEL) [6], have obtained promising results in the ﬁeld.”
A Survey of Sequential Pattern Mining (2017, methodology): “Moreover, sequential pattern mining can also be applied to time series (e.g. stock data), when discretization is performed as a pre-processing step [66] Sequential pattern mining is a very active research topic, where hundreds of papers present new algorithms and applications each year, including…”
Anomaly Detection for Discrete Sequences: A Survey (2012, methodology): “Note that these techniques were originally presented in the context of time series data, but can be extended to discrete sequences by discretizing the time series using techniques such as Symbolic ApproXimation (SAX) [69].”
Using dynamic time warping distances as features for improved time series classification (2016, methodology): “Symbolic aggregate approximation (SAX) is a method of representing time series in a lower dimensional space of symbolic " words " (Lin et al. 2007).”
Predicting ICU Mortality Risk by Grouping Temporal Trends from a Multivariate Panel of Physiologic Measurements (2016, methodology): “The NMF solver we used is the projected gradient NMF (Lin 2007) implemented in Scikit-learn (Pedregosa et al. 2011).”
TSclust: An R Package for Time Series Clustering (2014, methodology): “Equation 3 assumes that T/ω is an integer, but this assumption can be relaxed by weighting the contribution of data points placed in adjacent segments (see details on this generalization in Section 3.5 of Lin, Keogh, Wei, and Lonardi 2007).”
Cache locality is not enough: High-Performance Nearest Neighbor Search with Product Quantization Fast Scan (2015, positive): “Compression schemes, either generic [23, 12, 3, 25] or speciﬁc (e.g., SAX for time series [18]), have been widely adopted in database systems.”
Time-series clustering - A decade review (2015, methodology): “Some of similarity measures are proposed based on a specific time-series representations, for example, MINDIST which is compatible with SAX [84], and some of them work regardless of representation methods, or are compatible with raw time-series.”
Time series classification with ensembles of elastic distance measures (2015, positive): “Lin et al. (2012) propose a bag-of-patterns (BoP) approach that involves converting a time series into a discrete series using Symbolic Aggregate approximation (SAX) (Lin et al. 2007), creating a set of SAX words for each series through the application of a short sliding window, then using the frequ”
A Multivariate Timeseries Modeling Approach to Severity of Illness Assessment and Forecasting in ICU with Sparse, Heterogeneous Clinical Data (2015, background): “Another approach is time-series symbolization, which involves discretizing timeseries into sequences of symbols and attaching meaning to the groupings of the symbols (Lin et al. 2007; Saeed and Mark 2006; Syed and Guttag 2011).”
Visual Methods for Analyzing Time-Oriented Data (2008, background): “Using data abstraction is a current research topic more than ever [29].”
Toeplitz Inverse Covariance-Based Clustering of Multivariate Time Series Data (2017, methodology): “Methods include several variations of dynamic time warping [3, 23, 25, 39], symbolic representations [29, 30], and rule-based motif discovery [11, 28].”