A symbolic representation of time series, with implications for streaming algorithms

Lin, Jessica; Keogh, Eamonn; Lonardi, Stefano; Chiu, Bill

doi:10.1145/882085.882086

Cited by 725 publications

(567 citation statements)

References 0 publications

Supporting

Mentioning

541

Contrasting

Unclassified

Order By: Relevance

“…It consists of two "for" cycles (lines 5-17 and 18-30, respectively), which allow one to enter the index from its root, and move towards the leaves, first vertically (i.e., according to the time granularity, see lines [5][6][7][8][9][10][11][12][13][14][15][16][17], and then horizontally (i.e., according to the taxonomy of symbols, see lines [18][19][20][21][22][23][24][25][26][27][28][29][30]. The descent towards the index leaves can stop: (1) because the query time granularity/symbol taxonomy level is reached (lines 16 and 29, respectively, which update the output set at level nodes), or (2) because the index leaves are reached, but the index is incomplete so that the query time granularity/symbol taxonomy level is too specific and is not represented in the index itself (lines 10 and 23, which update the output set higher nodes).…”

Section: Examplementioning

confidence: 99%

“…Specifically, such strings are generated through a call to the build procedure, and saved in stringdummyset (line 7). Then, every string in stringdummyset is concatenated to all the elements already saved in out (lines [12][13][14][15][16][17][18][19]. On the other hand, if the current element e is a string (i.e., it is in the form SU B j , see Section 3.2), it is concatenated to all the elements already saved in out (lines [23][24][25][26][27][28][29][30][31][32][33][34][35].…”

Section: Examplementioning

confidence: 99%

See 1 more Smart Citation

A time series retrieval tool for sub-series matching

et al. 2015

View full text Add to dashboard Cite

The problem of retrieving time series similar to a specified query pattern has been recently addressed within the case based reasoning (CBR) literature. Providing a flexible and efficient way of dealing with such an issue is of paramount importance in many domains (e.g., medical), where the evolution of specific parameters is collected in the form of time series. In the past, we have developed a framework for retrieving time series, applying temporal abstractions. With respect to more classical (mathematical) approaches, our framework provides significant advantages. In particular, multi-level abstraction mechanisms and proper indexing techniques allow for flexible query issuing, and for efficient and interactive query answering. In this paper, we present an extension to such a framework, which aims to support sub-series matching as well. Indeed, sub-series retrieval may be crucial when the whole time series evolution is not of interest, while critical patterns to be searched for are only "local". Moreover, sometimes the relative order of patterns, but not their precise location in time, may be known. Finally, an interactive search, at different abstraction levels, may be required by the decision maker. Our extended framework (which is currently being applied in haemodialysis, but is domain independent) deals with all these issues.

show abstract

Section: Examplementioning

confidence: 99%

Section: Examplementioning

confidence: 99%

A time series retrieval tool for sub-series matching

et al. 2015

View full text Add to dashboard Cite

show abstract

“…This is the method used in SAX [51,52], which transforms a sequence of real numbers into a strings of characters, that can be classified using standard classification methods. Support vector machines have been successfully applied to time series data mining.…”

Section: Time Series Classificationmentioning

confidence: 99%

“…This approach extends SAX [51,52] adding new strings symbols. This representation is used for the task of classification of multivariate time series.…”

Section: Multivariate Time Series Classificationmentioning

confidence: 99%

Stacking for multivariate time series classification

Prieto

Alonso-González

Rodrı́guez

2013

Pattern Anal Applic

View full text Add to dashboard Cite

This work presents a novel approach to multivariate time series classification. The method exploits the multivariate structure of the time series and the possibilities of the stacking ensemble method. The basics of the method may be described in three steps: first, decomposing the multivariate time series on its constituent univariate time series; second, inducing a classifier for each univariate time series plus and additional multivariate classifier for the whole time series; third, creating the final multivariate time series classifier stacking the previous classifiers. The ensemble obtained has the potential to improve the accuracy of the single multivariate time series classifier. Several configurations of the stacking method have been tested on seven multivariate time series data sets. In five out of seven data sets, the proposed method obtains the smallest error rate. Moreover, in two out of seven data sets, stacking only the univariate time series classifiers provides the best results. The experimental results show that when a multivariate time series method does not produce an accurate classifier, stacking it with univariate time series classifiers is an alternative worthy of consideration.

show abstract

“…In References [17,18], the authors describe a method for transforming real time series values into symbolic representations (the time axe is divided into intervals with the same length, and a symbol is given to each interval). This new representation was used for searching motifs in each parameter separately using on a probabilistic approach.…”

Section: 42mentioning

confidence: 99%

Multi‐level temporal abstraction for medical scenario construction

Silvent

Dojat

Garbay

2004

Adaptive Control & Signal

View full text Add to dashboard Cite

The automatic recognition of typical pattern sequences (scenarios), as they are developing, is of crucial importance for computer-aided patient supervision. However, the construction of such scenarios directly from medical expertise is unrealistic in practice. In this paper, we present a methodology for data abstraction and for the extraction of specific events (data mining) to eventually construct such scenarios. Data abstraction and data mining are based on the management of three key concepts, data, information and knowledge, which are instantiated via an ontology specific of our medical domain application. After a detailed description of the proposed methodology, we apply it to the supervision of patients hospitalized in intensive care units. We report the results obtained for the extraction of typical abstracted pattern sequences during the process of weaning from mechanical ventilation.

show abstract

A symbolic representation of time series, with implications for streaming algorithms

Cited by 725 publications

References 0 publications

A time series retrieval tool for sub-series matching

A time series retrieval tool for sub-series matching

Stacking for multivariate time series classification

Multi‐level temporal abstraction for medical scenario construction

Contact Info

Product

Resources

About