Modeling Extreme Events in Time Series Prediction

Ding, Daizong; Zhang, Mi; Pan, Xudong; Yang, Min; He, Xiangnan

doi:10.1145/3292500.3330896

Cited by 108 publications

(54 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…the distribution calculated by GEV even with a high dispersion for highest return period. Even if it is well known that deep learning-based methods may results in weak performance for extreme events (Zhang et al, 2019), the results obtained here…”

Section: Coupled Modelling Approach: Strength and Uncertaintiescontrasting

confidence: 54%

Predicting Sediment Discharge at Water Treatment Plant Under Different Land Use Scenarios Coupling Expert-Based GIS Model and Deep Neural Network

Patault

Landemaine

Ledun

et al. 2020

Preprint

View full text Add to dashboard Cite

Abstract. Excessive sediment discharge at karstic springs and thus, water treatment plants, can be highly disruptive. It is essential for catchment stakeholders and drinking water supplier to reduce the impact of sediment on potable water supply, but their strategic choices must be based on simulations, integrating surface and groundwater transfers, and taking into account possible changes in land use. Karstic environments are particularly challenging as they face a lack of accurate physical description for the modelling process, and they can be seen as a black-box due to the non-linearity of the processes generating sediment discharge. The aim of the study was to assess the sediment discharge variability at a water treatment plant according to multiple realistic land use scenarios. To reach that goal, we developed a new coupled modelling approach with an erosion-runoff GIS model (WaterSed) and a deep neural network. The model was used in the Radicatel catchment (106 km2 in Normandy, France) where karstic spring water is extracted to a water treatment plant. The sediment discharge was simulated for five designed storm projects under current land use and compared to three land use scenarios (baseline, ploughing up of grassland, eco-engineering, best farming practices). Daily rainfall time series and WaterSed modelling outputs extracted at connected sinkholes were used as input data for the deep neural network model. The model structure was found by a classical trial and error procedure, and the model was trained on two significant hydrologic years. Evaluation on a test set showed a good performance of the model (NSE = 0.82), and the application of a monthly-backward chaining nested cross validation revealed that the model is able to generalize on new datasets. Simulations made for the three land use scenarios suggested that ploughing up 33 % of grasslands would not increase significantly sediment discharge at the water treatment plant (5 % in average). In the opposite, eco-engineering and best farming practices will significantly reduce sediment discharge at the water treatment plant (respectively in the range of 10–44 and 24–61 %). The coupling of these two strategies is the most efficient since it affects the hydro-sedimentary production and transfer processes (decreasing sediment discharge from 40 to 80 %). The coupled modelling approach developed in this study offers interesting opportunities for sediment discharge prediction at karstic springs or water treatment plant under multiple land use scenarios. It also provides robust decision-making tools for land use planning and drinking water suppliers.

show abstract

Section: Coupled Modelling Approach: Strength and Uncertaintiescontrasting

confidence: 54%

Predicting Sediment Discharge at Water Treatment Plant Under Different Land Use Scenarios Coupling Expert-Based GIS Model and Deep Neural Network

Patault

Landemaine

Ledun

et al. 2020

Preprint

View full text Add to dashboard Cite

show abstract

“…Recent contributions address the topic of extreme values, such as the work of Siffer et al (2017), Ding et al (2019) and Wang et al (2019). Based on the formalisation or use of tools from extreme value theory (the first two references) or combinations of pairwise preference classification and ordinal ranking, they illustrate well the issues tackled in this paper.…”

Section: Recent Workmentioning

confidence: 99%

“…Siffer et al (2017) use an error rate based on the quantiles of the distribution and a ROC analysis on discretised values. Ding et al (2019) use the Root Mean Squared Error and the F-Score with discretised values. Wang et al (2019) use statistical distance metrics for distributions and ROC analysis on discretised values.…”

Section: Recent Workmentioning

confidence: 99%

Imbalanced regression and extreme value prediction

Ribeiro

Moniz

2020

Mach Learn

View full text Add to dashboard Cite

Research in imbalanced domain learning has almost exclusively focused on solving classification tasks for accurate prediction of cases labelled with a rare class. Approaches for addressing such problems in regression tasks are still scarce due to two main factors. First, standard regression tasks assume each domain value as equally important. Second, standard evaluation metrics focus on assessing the performance of models on the most common values of data distributions. In this paper, we present an approach to tackle imbalanced regression tasks where the objective is to predict extreme (rare) values. We propose an approach to formalise such tasks and to optimise/evaluate predictive models, overcoming the factors mentioned and issues in related work. We present an automatic and non-parametric method to obtain relevance functions, building on the concept of relevance as the mapping of target values into non-uniform domain preferences. Then, we propose SERA, a new evaluation metric capable of assessing the effectiveness and of optimising models towards the prediction of extreme values while penalising severe model bias. An experimental study demonstrates how SERA provides valid and useful insights into the performance of models in imbalanced regression tasks.

show abstract

“…Gated recurrent units (GRU) were first proposed by [49] as a simplified and efficient modification of traditional RNN and LSTM cells. GRU models have successfully been applied to sequence modelling problems in the past [50,51], which makes them good candidates as a benchmark model for our paper. Unlike LSTM, GRU only has two gates, the update gate and the reset gate; since there is no output gate, GRU has no control over the memory content of the unit.…”

Section: Long-term Short-term Memory and Gated Recurrent Unitmentioning

confidence: 99%

Multi-Step Short-Term Wind Speed Prediction Using a Residual Dilated Causal Convolutional Network with Nonlinear Attention

Kumar

Tzou

2020

Energies

View full text Add to dashboard Cite

Wind energy is the most used renewable energy worldwide second only to hydropower. However, the stochastic nature of wind speed makes it harder for wind farms to manage the future power production and maintenance schedules efficiently. Many wind speed prediction models exist that focus on advance neural networks and/or preprocessing techniques to improve the accuracy. Since most of these models require a large amount of historic wind data and are validated using the data split method, the application to real-world scenarios cannot be determined. In this paper, we present a multi-step univariate prediction model for wind speed data inspired by the residual U-net architecture of the convolutional neural network (CNN). We propose a residual dilated causal convolutional neural network (Res-DCCNN) with nonlinear attention for multi-step-ahead wind speed forecasting. Our model can outperform long-term short-term memory networks (LSTM), gated recurrent units (GRU), and Res-DCCNN using sliding window validation techniques for 50-step-ahead wind speed prediction. We tested the performance of the proposed model on six real-world wind speed datasets with different probability distributions to confirm its effectiveness, and using several error metrics, we demonstrated that our proposed model was robust, precise, and applicable to real-world cases.

show abstract

Modeling Extreme Events in Time Series Prediction

Cited by 108 publications

References 26 publications

Predicting Sediment Discharge at Water Treatment Plant Under Different Land Use Scenarios Coupling Expert-Based GIS Model and Deep Neural Network

Predicting Sediment Discharge at Water Treatment Plant Under Different Land Use Scenarios Coupling Expert-Based GIS Model and Deep Neural Network

Imbalanced regression and extreme value prediction

Multi-Step Short-Term Wind Speed Prediction Using a Residual Dilated Causal Convolutional Network with Nonlinear Attention

Contact Info

Product

Resources

About