Reconstructing missing data sequences in multivariate time series: an application to environmental data

Parrella, Maria Lucia; Albano, Giuseppina; Rocca, Michele La; Perna, Cira

doi:10.1007/s10260-018-00435-9

Cited by 7 publications

(2 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One approach is autoregressive (AR) modelling. Liu and Molenaar [12] introduced vector autoregressive (VAR) models with one-step-ahead predictions, and Parrella, et al [13] applied spatial-dynamic autoregressive models to impute missing values across a cluster of air pollution monitors. To better capture the complex distributions found in multivariate times series new machine-learning (ML) based approaches, such as generative adversarial networks (GAN) and recurrent neural networks (RNN) show promise [14,15].…”

Section: Related Work On Time Series Imputationmentioning

confidence: 99%

Strategies for Imputation of High-Resolution Environmental Data in Clinical Randomized Controlled Trials

Kim

Kelly

Krishnan

et al. 2022

IJERPH

View full text Add to dashboard Cite

Time series data collected in clinical trials can have varying degrees of missingness, adding challenges during statistical analyses. An additional layer of complexity is introduced for missing data in randomized controlled trials (RCT), where researchers must remain blinded between intervention and control groups. Such restriction severely limits the applicability of conventional imputation methods that would utilize other participants’ data for improved performance. This paper explores and compares various methods to impute high-resolution temperature logger data in RCT settings. In addition to the conventional non-parametric approaches, we propose a spline regression (SR) approach that captures the dynamics of indoor temperature by time of day that is unique to each participant. We investigate how the inclusion of external temperature and energy use can improve the model performance. Results show that SR imputation results in 16% smaller root mean squared error (RMSE) compared to conventional imputation methods, with the gap widening to 22% when more than half of data is missing. The SR method is particularly useful in cases where missingness occurs simultaneously for multiple participants, such as concurrent battery failures. We demonstrate how proper modelling of periodic dynamics can lead to significantly improved imputation performance, even with limited data.

show abstract

Section: Related Work On Time Series Imputationmentioning

confidence: 99%

Strategies for Imputation of High-Resolution Environmental Data in Clinical Randomized Controlled Trials

Kim

Kelly

Krishnan

et al. 2022

IJERPH

View full text Add to dashboard Cite

show abstract

“…Different approaches have been proposed to address the missing data problem in multivariate time series, ranging from simple methods based on linear combinations of the neighbor contemporary observations [9] to state-of-the-art machine learning and deep learning methods [10][11][12], which are able to extract further information from data to gain insights about the process. Missing data in multivariate time series has been tackled in diverse domains, but, due to the criticality of process monitoring, it has attracted the attention of many researchers in the industry, specially in the field of energy systems and electricity consumption [5,13].…”

Section: Introductionmentioning

confidence: 99%

Gap imputation in related multivariate time series through recurrent neural network-based denoising autoencoder

Alonso,

Morán,

Pérez

et al. 2024

ICA

View full text Add to dashboard Cite

Technological advances in industry have made it possible to install many connected sensors, generating a great amount of observations at high rate. The advent of Industry 4.0 requires analysis capabilities of heterogeneous data in form of related multivariate time series. However, missing data can degrade processing and lead to bias and misunderstandings or even wrong decision-making. In this paper, a recurrent neural network-based denoising autoencoder is proposed for gap imputation in related multivariate time series, i.e., series that exhibit spatio-temporal correlations. The denoising autoencoder (DAE) is able to reproduce input missing data by learning to remove intentionally added gaps, while the recurrent neural network (RNN) captures temporal patterns and relationships among variables. For that reason, different unidirectional (simple RNN, GRU, LSTM) and bidirectional (BiSRNN, BiGRU, BiLSTM) architectures are compared with each other and to state-of-the-art methods using three different datasets in the experiments. The implementation with BiGRU layers outperforms the others, effectively filling gaps with a low reconstruction error. The use of this approach is appropriate for complex scenarios where several variables contain long gaps. However, extreme scenarios with very short gaps in one variable or no available data should be avoided.

show abstract

Bootstrap joint prediction regions for sequences of missing values in spatio-temporal datasets

et al. 2021

View full text Add to dashboard Cite

Missing data reconstruction is a critical step in the analysis and mining of spatio-temporal data. However, few studies comprehensively consider missing data patterns, sample selection and spatio-temporal relationships. To take into account the uncertainty in the point forecast, some prediction intervals may be of interest. In particular, for (possibly long) missing sequences of consecutive time points, joint prediction regions are desirable. In this paper we propose a bootstrap resampling scheme to construct joint prediction regions that approximately contain missing paths of a time components in a spatio-temporal framework, with global probability $$1-\alpha $$ 1 - α . In many applications, considering the coverage of the whole missing sample-path might appear too restrictive. To perceive more informative inference, we also derive smaller joint prediction regions that only contain all elements of missing paths up to a small number k of them with probability $$1-\alpha $$ 1 - α . A simulation experiment is performed to validate the empirical performance of the proposed joint bootstrap prediction and to compare it with some alternative procedures based on a simple nominal coverage correction, loosely inspired by the Bonferroni approach, which are expected to work well standard scenarios.

show abstract

Reconstructing missing data sequences in multivariate time series: an application to environmental data

Cited by 7 publications

References 18 publications

Strategies for Imputation of High-Resolution Environmental Data in Clinical Randomized Controlled Trials

Strategies for Imputation of High-Resolution Environmental Data in Clinical Randomized Controlled Trials

Gap imputation in related multivariate time series through recurrent neural network-based denoising autoencoder

Bootstrap joint prediction regions for sequences of missing values in spatio-temporal datasets

Contact Info

Product

Resources

About