Composite quantile regression neural network with applications

Xu, Qifa; Deng, Kai; Jiang, Cuixia; Sun, Fang; Huang, Xishi

doi:10.1016/j.eswa.2017.01.054

Cited by 80 publications

(51 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, partial monotonicity constraints are removed from the s-covariate; the exponential function is no longer applied to the relevant elements in W ðhÞ and all elements of w. The resulting model provides estimates of multiple regression quantiles, but crossing can now occur. This differs from the CQRNN model of Xu et al (2017), which estimates a single regression equation using the composite QR cost function, and MCQRNN, which additionally guarantees non-crossing of the multiple regression quantiles. Differences between the three models are illustrated in Fig.…”

Section: Monte Carlo Simulationmentioning

confidence: 99%

“…To date, the simultaneous estimation of multiple s-quantiles with guaranteed noncrossing has not been possible for QRNN models. However, simultaneous estimates for multiple values of s are used in the composite QRNN (CQRNN) model proposed by Xu et al (2017). CQRNN shares the same goal as the linear composite quantile regression (CQR) model (Zou and Yuan 2008), namely to borrow strength across multiple regression quantiles to improve the estimate of the true, unknown relationship between the covariates and the response.…”

Section: Monotone Composite Quantile Regression Neural Network (Mcqrnn)mentioning

confidence: 99%

“…Following Xu et al (2017), the number of hidden layer outputs in all models is set to J ¼ 4 for example 1 and J ¼ 5 for examples 2 and 3; for sake of simplicity, no weight penalty terms are added when fitting any of the models. (When comparing results with those reported by Xu et al (2017), note that omitting weight penalty regularization here leads to smaller inter-model differences in performance within both the training and testing samples, which suggests potential instability in hyperparameter selection in the previous study.) The goal is to estimate the true functional relationship specified by Eqs.…”

Section: Monte Carlo Simulationmentioning

confidence: 99%

“…Given the close relationship between the MCQRNN and CQRNN models, performance is first assessed via Monte Carlo simulation using the experimental setup adopted by Xu et al (2017) for CQRNN. The MCQRNN model is compared with standard MLP, QRNN, and CQRNN models on datasets generated for three example functions:…”

Section: Monte Carlo Simulationmentioning

confidence: 99%

“…In Sect. 3, the MCQRNN model is compared via Monte Carlo simulation to standard MLP, QRNN, and CQRNN models using combinations of three functions and error distributions from Xu et al (2017). In Sect.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Non-crossing nonlinear regression quantiles by monotone composite quantile regression neural network, with application to rainfall extremes

Cannon

2018

Stoch Environ Res Risk Assess

102

View full text Add to dashboard Cite

The goal of quantile regression is to estimate conditional quantiles for specified values of quantile probability using linear or nonlinear regression equations. These estimates are prone to ''quantile crossing'', where regression predictions for different quantile probabilities do not increase as probability increases. In the context of the environmental sciences, this could, for example, lead to estimates of the magnitude of a 10-year return period rainstorm that exceed the 20-year storm, or similar nonphysical results. This problem, as well as the potential for overfitting, is exacerbated for small to moderate sample sizes and for nonlinear quantile regression models. As a remedy, this study introduces a novel nonlinear quantile regression model, the monotone composite quantile regression neural network (MCQRNN), that (1) simultaneously estimates multiple non-crossing, nonlinear conditional quantile functions; (2) allows for optional monotonicity, positivity/ non-negativity, and generalized additive model constraints; and (3) can be adapted to estimate standard least-squares regression and non-crossing expectile regression functions. First, the MCQRNN model is evaluated on synthetic data from multiple functions and error distributions using Monte Carlo simulations. MCQRNN outperforms the benchmark models, especially for non-normal error distributions. Next, the MCQRNN model is applied to real-world climate data by estimating rainfall Intensity-Duration-Frequency (IDF) curves at locations in Canada. IDF curves summarize the relationship between the intensity and occurrence frequency of extreme rainfall over storm durations ranging from minutes to a day. Because annual maximum rainfall intensity is a non-negative quantity that should increase monotonically as the occurrence frequency and storm duration decrease, monotonicity and non-negativity constraints are key constraints in IDF curve estimation. In comparison to standard QRNN models, the ability of the MCQRNN model to incorporate these constraints, in addition to non-crossing, leads to more robust and realistic estimates of extreme rainfall.

show abstract

Section: Monte Carlo Simulationmentioning

confidence: 99%

Section: Monotone Composite Quantile Regression Neural Network (Mcqrnn)mentioning

confidence: 99%

Section: Monte Carlo Simulationmentioning

confidence: 99%

Section: Monte Carlo Simulationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Non-crossing nonlinear regression quantiles by monotone composite quantile regression neural network, with application to rainfall extremes

Cannon

2018

Stoch Environ Res Risk Assess

102

View full text Add to dashboard Cite

show abstract

Quantile regression neural network‐based fault detection scheme for wind turbines with application to monitoring a bearing

Fan

Jia³

et al. 2019

Wind Energy

Self Cite

View full text Add to dashboard Cite

Under the framework of normal behavior modeling, this paper develops a novel scheme for fault detection via quantile regression neural networks (QRNNs). The QRNN model is a combination of quantile regressions and neural networks. It is able to identify the normal status or extract the normal behavior data accurately and quickly through lower and upper regression quantiles.Additionally, it is flexible to explore the potential nonlinear patterns contained in the normal status by taking advantage of neural networks. Finally, we monitor the residuals produced from QRNN to detect faults by using the exponentially weighted moving average (EWMA) control chart. The utility of our scheme is illustrated by empirical analyses of bearing fault detection based on Supervisory Control and Data Acquisition (SCADA) data from a wind turbine. We find that the QRNN model outperforms the multiple linear regression (MLR) and back propagation neural networks (BPNNs) in terms of mean absolute error (MAE). Besides, the obtained relationship between the width of control limits in EWMA and the number of alarms provides an important and convenient way for practical applications. KEYWORDS exponentially weighted moving average, fault detection, normal behavior modeling, quantile regression neural networks, wind turbine INTRODUCTIONWind energy is one of the most significant clean renewable energies for electrical generation. An increasing number of wind farms have been built to fulfill human's growing power demand. However, wind turbines (WTs) are typically deployed in a remote or offshore location with abundant wind energy, where performing regular inspections to maintain normal operations will cost a lot. In addition, unscheduled and reactive maintenances could increase the meaningless downtime, causing the loss of revenue. Therefore, how to perform more effective operations and maintenance procedures via monitoring remotely with lower cost has attracted more and more attentions from both academics and practitioners.Condition monitoring systems (CMSs) require not only high-level knowledge about the systems but also the high-cost vibration sensors, which hinders from the prevalence of CMSs. 1-3 On the contrary, the data collected via Supervisory Control and Data Acquisition (SCADA) systems that archive useful information in a convenient manner is universally available. Extensive efforts have been devoted to developing various methods for fault detection of WTs based on SCADA data.Based on SCADA data, the residuals produced from three models including a generalized mapping regressor (GMR), a general regression neural network (GRNN), and a feed-forward multilayer perceptron (MLP) have already been used to monitor the power curve of WTs. 4 The copula approach, estimating bivariate probability distribution function, is proposed to model the power curve so that the deviation between the existing and expected behavior can be checked. 5 The Mahalanobis distance is applied to detect faults in the optional curves containing the power curve, rotor curv...

show abstract

Composite Quantile Regression Long Short-Term Memory Network

Xie

Wen

2019

Artificial Neural Networks and Machine Learning – ICANN 2019: Text and Time Series

View full text Add to dashboard Cite

Composite quantile regression neural network with applications

Cited by 80 publications

References 26 publications

Non-crossing nonlinear regression quantiles by monotone composite quantile regression neural network, with application to rainfall extremes

Non-crossing nonlinear regression quantiles by monotone composite quantile regression neural network, with application to rainfall extremes

Quantile regression neural network‐based fault detection scheme for wind turbines with application to monitoring a bearing

Composite Quantile Regression Long Short-Term Memory Network

Contact Info

Product

Resources

About