VS-GRU: A Variable Sensitive Gated Recurrent Neural Network for Multivariate Time Series with Massive Missing Values

Li, Qianting; Xu, Yong

doi:10.3390/app9153041

Cited by 30 publications

(15 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…An extension to missing indicators, capable of incorporating both IP and IO, is to derive a summary of the measurement process and include this as a predictor. 15 , 24 , 31–44 Examples include a count of the number of measurements (eg, throughout a critical care admission), 37 weighted counts, 42 combined missing indicators, 31 missingness rates over time, 32 time intervals between measures, 33–35 embedding vectors that represent missing values, 36 or information relating to hospital processes. 38 , 39 …”

Section: Resultsmentioning

confidence: 99%

Informative presence and observation in routine health data: A review of methodology for clinical risk prediction

Sisk

Lin

Sperrin

et al. 2020

Journal of the American Medical Informatics Association

View full text Add to dashboard Cite

Objective Informative presence (IP) is the phenomenon whereby the presence or absence of patient data is potentially informative with respect to their health condition, with informative observation (IO) being the longitudinal equivalent. These phenomena predominantly exist within routinely collected healthcare data, in which data collection is driven by the clinical requirements of patients and clinicians. The extent to which IP and IO are considered when using such data to develop clinical prediction models (CPMs) is unknown, as is the existing methodology aiming at handling these issues. This review aims to synthesize such existing methodology, thereby helping identify an agenda for future methodological work. Materials and Methods A systematic literature search was conducted by 2 independent reviewers using prespecified keywords. Results Thirty-six articles were included. We categorized the methods presented within as derived predictors (including some representation of the measurement process as a predictor in the model), modeling under IP, and latent structures. Including missing indicators or summary measures as predictors is the most commonly presented approach amongst the included studies (24 of 36 articles). Discussion This is the first review to collate the literature in this area under a prediction framework. A considerable body relevant of literature exists, and we present ways in which the described methods could be developed further. Guidance is required for specifying the conditions under which each method should be used to enable applied prediction modelers to use these methods. Conclusions A growing recognition of IP and IO exists within the literature, and methodology is increasingly becoming available to leverage these phenomena for prediction purposes. IP and IO should be approached differently in a prediction context than when the primary goal is explanation. The work included in this review has demonstrated theoretical and empirical benefits of incorporating IP and IO, and therefore we recommend that applied health researchers consider incorporating these methods in their work.

show abstract

Section: Resultsmentioning

confidence: 99%

Informative presence and observation in routine health data: A review of methodology for clinical risk prediction

Sisk

Lin

Sperrin

et al. 2020

Journal of the American Medical Informatics Association

View full text Add to dashboard Cite

show abstract

“…Although in the latter works the missing rate of variables were huge (more than 70%) and different from each other, the model behaved the same for all variables. In [21], the missing rate of each variable was included in the inputs of GRU to reduce the harmful impact of variables with high missing rate on variables with low missing rate. Considering the fact that the mentioned literature of MNAR category was all analyzed over medical datasets, the work in [38] extended the analysis over the computer vision domain as well.…”

Section: Related Workmentioning

confidence: 99%

“…Taking into account recurrent methods, the variants of Recurrent Neural Network (RNN) like Gated Recurrent Unit (GRU) [19][20][21] and Long Short-Term Memory (LSTM) [22][23][24] have demonstrated promising results in extracting temporal features and some could deal with huge missingness as well. It is notable that the approaches in [19][20][21][22] are among few works on addressing massive MNAR missingness in MTS (more than 70%). However, the main focus of these works have been on the classification task of medical domains.…”

Section: Introductionmentioning

confidence: 99%

“…Exploring massive MNAR missingness in the regression task of other domains is still an open research area. More importantly, since not all variables follow the same missing rate in many applications, it is worth investigating the missing rate of each variable along with other missing information in the recurrent methods [21]. To the best of our knowledge, this has not yet been explored in an LSTM-based structure, having the power of controlling the information flow with its memory unit, jointly with other missing information.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Novel LSTM for Multivariate Time Series with Massive Missingness

Fouladgar

Främling

2020

Sensors

View full text Add to dashboard Cite

Multivariate time series with missing data is ubiquitous when the streaming data is collected by sensors or any other recording instruments. For instance, the outdoor sensors gathering different meteorological variables may encounter low material sensitivity to specific situations, leading to incomplete information gathering. This is problematic in time series prediction with massive missingness and different missing rate of variables. Contribution addressing this problem on the regression task of meteorological datasets by employing Long Short-Term Memory (LSTM), capable of controlling the information flow with its memory unit, is still missing. In this paper, we propose a novel model called forward and backward variable-sensitive LSTM (FBVS-LSTM) consisting of two decay mechanisms and some informative data. The model inputs are mainly the missing indicator, time intervals of missingness in both forward and backward direction and missing rate of each variable. We employ this information to address the so-called missing not at random (MNAR) mechanism. Separately learning the features of each parameter, the model becomes adapted to deal with massive missingness. We conduct our experiment on three real-world datasets for the air pollution forecasting. The results demonstrate that our model performed well along with other LSTM-derivation models in terms of prediction accuracy.

show abstract

“…The missing data completion has been applied to many fields [13][14][15][16]. In the field of dams, Lv et al [17] pointed out that the interpolation methods of observation data mainly include internal physical association interpolation and mathematical interpolation and introduced the principle and process of linear interpolation.…”

Section: Introduction 1literature Reviewsmentioning

confidence: 99%

A Completion Method for Missing Concrete Dam Deformation Monitoring Data Pieces

Wang

Zhu

et al. 2021

Applied Sciences

View full text Add to dashboard Cite

A concrete dam is an important water-retaining hydraulic structure that stops or restricts the flow of water or underground streams. It can be regarded as a constantly changing complex system. The deformation of a concrete dam can reflect its operation behaviors most directly among all the effect quantities. However, due to the change of the external environment, the failure of monitoring instruments, and the existence of human errors, the obtained deformation monitoring data usually miss pieces, and sometimes the missing pieces are so critical that the remaining data fail to fully reflect the actual deformation patterns. In this paper, the composition, characteristics, and contamination of the concrete dam deformation monitoring information are analyzed. From the single-value missing data completion method based on the nonlocal average method, a multi-value missing data completion method using BP (back propagation) mapping of spatial adjacent points is proposed to improve the accuracy of analysis and pattern prediction of concrete dam deformation behaviors. A case study is performed to validate the proposed method.

show abstract

VS-GRU: A Variable Sensitive Gated Recurrent Neural Network for Multivariate Time Series with Massive Missing Values

Cited by 30 publications

References 31 publications

Informative presence and observation in routine health data: A review of methodology for clinical risk prediction

Informative presence and observation in routine health data: A review of methodology for clinical risk prediction

A Novel LSTM for Multivariate Time Series with Massive Missingness

A Completion Method for Missing Concrete Dam Deformation Monitoring Data Pieces

Contact Info

Product

Resources

About