Clustering of time-series subsequences is meaningless: implications for previous and future research

Keogh, Eamonn; Lin, Jessica

doi:10.1007/s10115-004-0172-7

Cited by 331 publications

(176 citation statements)

References 26 publications

Supporting

Mentioning

158

Contrasting

Unclassified

Order By: Relevance

“…However, the naive implementation of subsequence clustering (SSC) using a sliding window and k-Means is controversial, as it is prone to producing undesirable and unpredictable results, as was previously demonstrated and analyzed in several publications, e.g. [4,1,3]. Indeed, within our strain data application, we notice some of the mentioned phenomena, although not all.…”

Section: Introductionmentioning

confidence: 60%

“…However, in a recent paper by Keogh et al [4], it was shown that despite the intuitive match, SSC is prone to a number of undesirable behavior that make it, in the view of the authors, unsuitable for the task at hand. Since then, a number of papers (e.g.…”

Section: Subsequence Clustering Equals Event Detection?mentioning

confidence: 99%

“…In order to model all the different kinds of traffic events represented in the strain data, we investigate the effectiveness of time series subsequence clustering [2,4,1,3], which essentially employs a sliding window technique to generate input for the chosen clustering method. However, the naive implementation of subsequence clustering (SSC) using a sliding window and k-Means is controversial, as it is prone to producing undesirable and unpredictable results, as was previously demonstrated and analyzed in several publications, e.g.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Traffic Events Modeling for Structural Health Monitoring

Vespier

Knobbe

Vanschoren

et al. 2011

Advances in Intelligent Data Analysis X

View full text Add to dashboard Cite

Abstract. Since 2008, a sensor network on a major Dutch highway bridge has been monitoring the structural health of the bridge, by measuring various parameters at different locations along the infrastructure. These parameters include strain, vibration and climate. The aim of the InfraWatch project is to model the health and behavior of the bridge by analyzing the large quantities of data that the sensors produce. One of the many forms of modeling involved is the identification of traffic events (cars, trucks, congestion and so on), as knowing when they occur, and of what nature they are, will enable modeling the response of the bridge to each of these events. In this paper, we approach the problem as a time series subsequence clustering problem. As it is known that such a clustering method can be problematic on certain types of time series, we verified known problems on the InfraWatch data. Indeed some of the undesired phenomena occurred in our case, but to a lesser extent than previously suggested. We introduce a new distance measure over subsequences that discourages the observed behavior and allows us to identify traffic events reliably, even on large quantities of data.

show abstract

Section: Introductionmentioning

confidence: 60%

Section: Subsequence Clustering Equals Event Detection?mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Traffic Events Modeling for Structural Health Monitoring

Vespier

Knobbe

Vanschoren

et al. 2011

Advances in Intelligent Data Analysis X

View full text Add to dashboard Cite

show abstract

“…STSC-based stream mining methods enjoyed popularity until a surprising fact was discovered in 2003 [8]: k-means STSC is "meaningless" as a pattern discovery technique in that the resultant cluster centers tend to form sinusoidal pseudo-patterns almost independently of the input time series.…”

Section: Introductionmentioning

confidence: 99%

“…The data mining and machine learning communities were surprised when Keogh et al (2003) pointed out that the k-means cluster centers in subsequence time-series clustering become sinusoidal pseudopatterns for almost all kinds of input time-series data. Understanding this mechanism is an important open problem in data mining.…”

mentioning

confidence: 99%

Why Does Subsequence Time-Series Clustering Produce Sine Waves?

Idé

2006

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. The data mining and machine learning communities were surprised when Keogh et al. (2003) pointed out that the k-means cluster centers in subsequence time-series clustering become sinusoidal pseudopatterns for almost all kinds of input time-series data. Understanding this mechanism is an important open problem in data mining. Our new theoretical approach (based on spectral clustering and translational symmetry) explains why the cluster centers of k-means naturally tend to form sinusoidal patterns.

show abstract

Clustering multivariate time‐series data

Singhal¹,

Seborg²

2005

Journal of Chemometrics

138

View full text Add to dashboard Cite

A new methodology for clustering multivariate time-series data is proposed. The new methodology is based on calculating the degree of similarity between multivariate time-series datasets using two similarity factors. One similarity factor is based on principal component analysis and the angles between the principal component subspaces while the other is based on the Mahalanobis distance between the datasets. The standard K-means clustering algorithm is modified to cluster multivariate time-series datasets using similarity factors. Simulation data from two nonlinear dynamic systems: a batch fermentation and a continuous exothermic chemical reactor, are clustered to demonstrate the effectiveness of the proposed technique. Comparisons with existing clustering methods show several advantages of the proposed method.

show abstract

Clustering of time-series subsequences is meaningless: implications for previous and future research

Cited by 331 publications

References 26 publications

Traffic Events Modeling for Structural Health Monitoring

Traffic Events Modeling for Structural Health Monitoring

Why Does Subsequence Time-Series Clustering Produce Sine Waves?

Clustering multivariate time‐series data

Contact Info

Product

Resources

About