Scaling up dynamic time warping for datamining applications

Keogh, Eamonn; Pazzani, Michael J.

doi:10.1145/347090.347153

Cited by 655 publications

(353 citation statements)

References 11 publications

Supporting

Mentioning

338

Contrasting

Unclassified

Order By: Relevance

“…The metric builds an optimal alignment of points among two time series by dynamically constructing a progressive cost matrix. It computes the the path of the minimal overall point pairs' distance [2]. Adding warping window size constraints have been reported to occasionally boost the classification [16].…”

Section: Time Series Classificationmentioning

confidence: 99%

“…The generality of interest is influenced by the large number of problems involving time series, ranging from financial econometrics up to medical diagnosis [1]. The most widely applied time series classifier is the nearest neighbor (NN) empowered with a similarity/distance metric called Dynamic Time Warping (DTW), hereafter jointly denoted as DTW-NN [2]. Due to the accuracy of DTW in detecting pattern variations, DTW-NN has been characterized as a hard-to-beat baseline [3].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Invariant Time-Series Classification

Grabocka

Νανόπουλος

Schmidt-Thieme

2012

Machine Learning and Knowledge Discovery in Databases

View full text Add to dashboard Cite

Abstract. Time-series classification is a field of machine learning that has attracted considerable focus during the recent decades. The large number of time-series application areas ranges from medical diagnosis up to financial econometrics. Support Vector Machines (SVMs) are reported to perform non-optimally in the domain of time series, because they suffer detecting similarities in the lack of abundant training instances. In this study we present a novel time-series transformation method which significantly improves the performance of SVMs. Our novel transformation method is used to enlarge the training set through creating new transformed instances from the support vector instances. The new transformed instances encapsulate the necessary intra-class variations required to redefine the maximum margin decision boundary. The proposed transformation method utilizes the variance distributions from the intra-class warping maps to build transformation fields, which are applied to series instances using the Moving Least Squares algorithm. Extensive experimentations on 35 time series datasets demonstrate the superiority of the proposed method compared to both the Dynamic Time Warping version of the Nearest Neighbor and the SVMs classifiers, outperforming them in the majority of the experiments.

show abstract

Section: Time Series Classificationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Invariant Time-Series Classification

Grabocka

Νανόπουλος

Schmidt-Thieme

2012

Machine Learning and Knowledge Discovery in Databases

View full text Add to dashboard Cite

show abstract

“…Recent data mining research has focused on using specialised similarity measures such as dynamic time warping (DTW) [5] in conjunction with lazy classifiers [4] to capture both correlation based and shape based similarity. DTW is is a natural generalisation of using Euclidean distance based methods and is often seen as a means of compensating against slight phase shift rather than capturing phase independent similarity.…”

Section: Time Series Classificationmentioning

confidence: 99%

Classification of Household Devices by Electricity Usage Profiles

Lines

Bagnall

Caiger-Smith

et al. 2011

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. This paper investigates how to classify household items such as televisions, kettles and refrigerators based only on their electricity usage profile every 15 minutes over a fixed interval of time. We address this time series classification problem through deriving a set of features that characterise the pattern of usage and the amount of power used when a device is on. We evaluate a wide range of classifiers on both the raw data and the derived feature set using both a daily and weekly usage profile and demonstrate that whilst some devices can be identified with a high degree of accuracy, others are very hard to disambiguate with this granularity of data.

show abstract

“…One of the most surprising recent results is, however, that the simple 1-nearest neighbor (1-NN) classifier using dynamic time warping (DTW) distance [15] has been shown to be competitive or superior to many state-of-the art time-series classification methods [6], [9], [13]. These results inspired intensive research of DTW in the last decade: this method has been examined in depth (for a thorough summary of results see [11]), while the improvements in its accuracy [2], [12] and efficiency [10] allowed to apply it to large, real-word recognition problems.…”

Section: Introductionmentioning

confidence: 99%

Fusion of Similarity Measures for Time Series Classification

Búza

Νανόπουλος

Schmidt-Thieme

2011

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. Time series classification, due to its applications in various domains, is one of the most important data-driven decision tasks of artificial intelligence. Recent results show that the simple nearest neighbor method with an appropriate distance measure performs surprisingly well, outperforming many state-of-the art methods. This suggests that the choice of distance measure is crucial for time series classification. In this paper we shortly review the most important distance measures of the literature, and, as major contribution, we propose a framework that allows fusion of these different similarity measures in a principled way. Within this framework, we develop a hybrid similarity measure. We evaluate it in context of time series classification on a large, publicly available collection of 35 real-world datasets and we show that our method achieves significant improvements in terms of classification accuracy.

show abstract

Scaling up dynamic time warping for datamining applications

Cited by 655 publications

References 11 publications

Invariant Time-Series Classification

Invariant Time-Series Classification

Classification of Household Devices by Electricity Usage Profiles

Fusion of Similarity Measures for Time Series Classification

Contact Info

Product

Resources

About