Mitigating Drift in Time Series Data with Noise Augmentation

Fields, Tonya; Hsieh, George; Chenou, Jules

doi:10.1109/csci49370.2019.00046

Cited by 32 publications

(27 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is able to do this by effectively creating new patterns with the assumption that the unseen test patterns are only different from the training patterns by a factor of noise. In addition, jittering has been shown to help mitigate time series drift for various neural network models [28]. Time series drift when the data distribution changes due to the introduction of new data.…”

Section: Magnitude Domain Transformationsmentioning

confidence: 99%

“…Similar to data augmentation for images, most data augmentation techniques for time series are based on random transformations of the training data. For example, adding random noise [28], slicing or cropping [29], scaling [30], random warping in the time dimension [28,30], and frequency warping [31]. Examples The problem with random transformation-based data augmentation is that there is a diverse amount of time series with each having different properties, and not every transformation is applicable to every dataset.…”

Section: Introductionmentioning

confidence: 99%

“…Similar to data augmentation for images, most data augmentation techniques for time series are based on random transformations of the training data. For example, adding random noise [19], slicing or cropping [20], scaling [21], random warping in the time dimension [19], [21], and frequency warping [22]. Examples of random transformationbased methods are shown in Fig.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

An empirical survey of data augmentation for time series classification with neural networks

2021

View full text Add to dashboard Cite

In recent times, deep artificial neural networks have achieved many successes in pattern recognition. Part of this success can be attributed to the reliance on big data to increase generalization. However, in the field of time series recognition, many datasets are often very small. One method of addressing this problem is through the use of data augmentation. In this paper, we survey data augmentation techniques for time series and their application to time series classification with neural networks. We propose a taxonomy and outline the four families in time series data augmentation, including transformation-based methods, pattern mixing, generative models, and decomposition methods. Furthermore, we empirically evaluate 12 time series data augmentation methods on 128 time series classification datasets with six different types of neural networks. Through the results, we are able to analyze the characteristics, advantages and disadvantages, and recommendations of each data augmentation method. This survey aims to help in the selection of time series data augmentation for neural network applications.

show abstract

Section: Magnitude Domain Transformationsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

“…Similar to data augmentation for images, most data augmentation techniques for time series are based on random transformations of the training data. For example, adding random noise [19], slicing or cropping [20], scaling [21], random warping in the time dimension [19], [21], and frequency warping [22]. Examples of random transformationbased methods are shown in Fig.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

An empirical survey of data augmentation for time series classification with neural networks

2021

View full text Add to dashboard Cite

show abstract

“…Furthermore, phenological events and the underlying trend of the VI curves of each of the crop are accurately modeled using computationally optimal transformations [64]. Unlike the existing transform-based approaches, the proposed approach does not alter the wavelet coefficients that characterize a given crop class [35], [45].…”

Section: B Improvement In Classification Resultsmentioning

confidence: 99%

“…The conventional approaches, such as geometric transformations, kernel filters (smoothing or enhancement), image mixing (interchanging slices), and random erasing (removing random slices), are suited only for image-related tasks [35], [40], [42]. For time-series data, augmentations such as jittering (random noise addition) [45], slicing (cropping) [46], magnitude warping (smooth element-wise magnitude change) [18], [38], permutation (rearranging slices) [45], [47], rotation (flipping for univariate; rotation for multivariate) [35], scaling (pattern-wise magnitude change) [47], random warping in the time dimension (time step deformation) [45], [47], and frequency warping (frequency deformation) [48] are adopted. The slicing-based augmentations responded positively with more extended time series, but pattern mixing methods are negatively correlated to the time series length.…”

mentioning

confidence: 99%

Augmentation of Vegetation Index Curves Considering the Crop-Specific Phenological Characteristics

Arun

Karnieli

2022

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

The state-of-the-art crop phenological classifiers use vegetation index (VI) time-series data and deep learning (DL) techniques. However, the scarcity of training samples limits the performance of these approaches. Unlike the conventional augmentation techniques, the data augmentation of VI curves should preserve the crop-specific phenological events. The DLbased augmentation approaches do not give good results when the training samples are limited. Also, the conventional approaches such as translation, rotation, scaling, and wrapping do not preserve the characteristic features of the index curves, thereby making them inappropriate for the VI-curve-based augmentations. This study proposes a non-DL-based data augmentation strategy that requires only a minimal number of actual training samples. In the proposed approach, the periodic phenological events and the underlying trend for each crop class are modeled to improve the augmentation. The trends of different crop classes are estimated by jointly maximizing the autocorrelation and variance, while the optimal subsequences are generalized as the phenological events. The proposed augmentation strategy of using Maximal Overlap Discrete Wavelet Transform (MODWT) for obtaining the surrogates that retain the crop-specific features and periodicities significantly improves the results. It may be noted that the proposed approach does not alter the wavelet coefficients that are characteristics of a given crop class. The experiments using time series VI data, covering 90 fields of wheat and 60 fields of barley, confirm better accuracy of the proposed augmentation approaches as compared to the prominent approaches.

show abstract

Novel statistical method for data drift detection in satellite telemetry

Praveen,

kuchhal,

Choudhury

2024

Int J Communication

View full text Add to dashboard Cite

SummaryAutonomy is becoming a prime requirement for satellite mission control operations. Data‐driven methods like Machine Learning models are playing a key role in bringing in autonomy. Health keeping data from satellite telemetry is a key ingredient in these data‐driven methods. In real‐world satellite operations, the health‐keeping telemetry data gradually drifts due to adverse space weather effects and wear and tear of electronic and mechanical components. The key question that arises is how to detect and quantify the data drift which is generally a gradual phenomenon. This paper discusses a novel statistical method for detecting data drift occurring in satellite telemetry. For the purpose of experimental work in this paper, an actual telemetry data set of the BUS CURRENT sensor which is part of the Electrical Power System of a Low Earth Orbit Satellite was considered. Data drift detection test was carried out using this sensor data using the developed novel statistical method and with Kolmogorov Smirnov test which is a probabilistic method. Both results are analysed and compared. Thereafter novel statical method was used to check its efficacy using a synthetic data set with induced drift.

show abstract

Mitigating Drift in Time Series Data with Noise Augmentation

Cited by 32 publications

References 3 publications

An empirical survey of data augmentation for time series classification with neural networks

An empirical survey of data augmentation for time series classification with neural networks

Augmentation of Vegetation Index Curves Considering the Crop-Specific Phenological Characteristics

Novel statistical method for data drift detection in satellite telemetry

Contact Info

Product

Resources

About