Efficient Time Series Classification under Template Matching Using Time Warping Alignment

Srisai, Dararat; Ratanamahatana, Chotirat Ann

doi:10.1109/iccit.2009.291

Cited by 6 publications

(8 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…1-NN classification algorithm with DTW seems to be the most widely used classifier; it was shown to be highly accurate [Xi et al 2006], though computing speed is significantly affected by repeated DTW computations. To overcome this limitation [Srisai and Ratanamahatana 2009] proposed a template construction algorithm based on the Accurate Shape Averaging (ASA) technique. Each training class is represented by only one sequence so that any incoming series is compared only with one averaged template per class.…”

Section: Classificationmentioning

confidence: 99%

Time-series data mining

2012

View full text Add to dashboard Cite

In almost every scientific field, measurements are performed over time. These observations lead to a collection of organized data called time series. The purpose of time series data mining is to try to extract all meaningful knowledge from the shape of data. Even if humans have a natural capacity to perform these tasks, it remains a complex problem for computers. In this paper we intend to provide a survey of the techniques applied for time series data mining. The first part is devoted to an overview of the tasks that have captured most of the interest of researchers. Considering that in most cases, time series task relies on the same components for implementation, we divide the literature depending on these common aspects, namely representation techniques, distance measures and indexing methods. The study of the relevant literature has been categorized for each individual aspects. Four types of robustness could then be formalized and any kind of distance could then be classified. Finally, the study submit various research trends and avenues that can be explored in the near future. We hope that this paper can provide a broad and deep understanding of the time series data mining research field.

show abstract

Section: Classificationmentioning

confidence: 99%

Time-series data mining

2012

View full text Add to dashboard Cite

show abstract

“…1. Unsupervised prototype generation [1,26,29,30,31,35,39,41]: Unsupervised methods cluster the training examples of every class separately. Centroids of the clusters are computed by averaging warped time series, which is non-trivial as compared to averaging vectors [27,36].…”

Section: Introductionmentioning

confidence: 99%

Asymmetric learning vector quantization for efficient nearest neighbor classification in dynamic time warping spaces

Jain

Schultz

2018

Pattern Recognition

View full text Add to dashboard Cite

The nearest neighbor method together with the dynamic time warping (DTW) distance is one of the most popular approaches in time series classification. This method suffers from high storage and computation requirements for large training sets. As a solution to both drawbacks, this article extends learning vector quantization (LVQ) from Euclidean spaces to DTW spaces. The proposed LVQ scheme uses asymmetric weighted averaging as update rule. Empirical results exhibited superior performance of asymmetric generalized LVQ (GLVQ) over other state-of-the-art prototype generation methods for nearest neighbor classification.

show abstract

“…The concept of template has been introduced to time series to detect specific patterns or shapes [18,19,40,41,48]. Frank et al [18] propose Geometric Template Matching (GeTeM) which uses time-delay embeddings for building models from segments of time series and compares the reconstructed dynamical systems in terms of their state space as well as their dynamics.…”

Section: Related Workmentioning

confidence: 99%

“…In [19], a novel and flexible approach is proposed based on segmental semi-Markov models. In [40,41,48], meaningful templates are constructed with shape-based averaging algorithms, such as Prioritized Shape Averaging (PSA) [40] and Accurate Shape Averaging (ASA) [48]. Wei et al propose the Atomic Wedgie method "that exploits the commonality among the predefined patterns to allow monitoring at higher bandwidths, while maintaining a guarantee of no false dismissals" [56].…”

Section: Related Workmentioning

confidence: 99%

“…When matching a predefined phenomenon (a template [40,41,48]) with the time series under investigation, it is not always required to involve every individual measurement in the selected interval and in the template. In fact, when a certain level of fuzzy matching is required, it makes sense to somehow simplify the signal, or extract some key features that are characteristic for the sequence in question.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Predefined pattern detection in large time series

Miao

Vespier

Cachucho

et al. 2016

Information Sciences

View full text Add to dashboard Cite

a b s t r a c tPredefined pattern detection from time series is an interesting and challenging task. In order to reduce its computational cost and increase effectiveness, a number of time series representation methods and similarity measures have been proposed. Most of the existing methods focus on full sequence matching, that is, sequences with clearly defined beginnings and endings, where all data points contribute to the match. These methods, however, do not account for temporal and magnitude deformations in the data and result to be ineffective on several real-world scenarios where noise and external phenomena introduce diversity in the class of patterns to be matched. In this paper, we present a novel pattern detection method, which is based on the notions of templates, landmarks, constraints and trust regions. We employ the Minimum Description Length (MDL) principle for time series preprocessing step, which helps to preserve all the prominent features and prevents the template from overfitting. Templates are provided by common users or domain experts, and represent interesting patterns we want to detect from time series. Instead of utilising templates to match all the potential subsequences in the time series, we translate the time series and templates into landmark sequences, and detect patterns from landmark sequence of the time series. Through defining constraints within the template landmark sequence, we effectively extract all the landmark subsequences from the time series landmark sequence, and obtain a number of landmark segments (time series subsequences or instances). We model each landmark segment through scaling the template in both temporal and magnitude dimensions. To suppress the influence of noise, we introduce the concept of trust region, which not only helps to achieve an improved instance model, but also helps to catch the accurate boundaries of instances of the given template. Based on the similarities derived from instance models, we introduce the probability density function to calculate a similarity threshold. The threshold can be used to judge if a landmark segment is a true instance of the given template or not. To evaluate the effectiveness and efficiency of the proposed method, we apply it to two real-world datasets. The results show that our method is capable of detecting patterns of temporal and magnitude deformations with competitive performance.

show abstract

Efficient Time Series Classification under Template Matching Using Time Warping Alignment

Cited by 6 publications

References 10 publications

Time-series data mining

Time-series data mining

Asymmetric learning vector quantization for efficient nearest neighbor classification in dynamic time warping spaces

Predefined pattern detection in large time series

Contact Info

Product

Resources

About