Recurrent Neural Networks for Time Series Forecasting: Current status and future directions

Hewamalage, Hansika; Bergmeir, Christoph; Bandara, Kasun

doi:10.1016/j.ijforecast.2020.06.008

Cited by 736 publications

(406 citation statements)

References 39 publications

Supporting

Mentioning

385

Contrasting

Unclassified

Order By: Relevance

“… Zeroual et al, 2020 , Tomar and Gupta, 2020 , Czarnowski et al, 2008 , El Zowalaty and Järhult, 2020 , Shahid et al, 2020 , Hewamalage et al, 2021 , Jiménez et al, 2020 , Kaushik et al, 2020 , BhedadJamshidi et al, 2020 , Ribeiro et al, 2020 , Naudé, 2020 , Arora et al, 2020 , Ribeiro et al, 2020 , Ogundokun et al, 2020 , Alzahrani et al, 2020 , Shastri et al, 2020 , Alakus and Turkoglu, 2020 , Papastefanopoulos et al, 2020 , Chimmula and Zhang, 2020 , Wang et al, 2020 , Wang et al, 2020 , DataGov , Car et al, 2020 .…”

Section: Uncited Referencesmentioning

confidence: 99%

Forecasting of COVID-19 cases using deep learning models: Is it reliable and practically significant?

Devaraj¹,

et al. 2021

View full text Add to dashboard Cite

Section: Uncited Referencesmentioning

confidence: 99%

Forecasting of COVID-19 cases using deep learning models: Is it reliable and practically significant?

Devaraj¹,

et al. 2021

View full text Add to dashboard Cite

“…Alternative NN architectures, known as recurrent neural networks (RNN), are also employed for forecasting, where connectivity information forming the input-output mapping is stored as hidden states. This RNN holds a recursive structure that implies that fewer parameters are demanded than NAR models to learn the relation input-output in data [58]. LSTM are architectures based on RNN, where the ability to preserve part of the information that belongs to the hidden layer can be used in specific moments for forecasting [59].…”

Section: Long Short Term Memory Modelmentioning

confidence: 99%

Data-Driven Trajectory Prediction of Grid Power Frequency Based on Neural Models

et al. 2021

View full text Add to dashboard Cite

Frequency in power systems is a real-time information that shows the balance between generation and demand. Good system frequency observation is vital for system security and protection. This paper analyses the system frequency response following disturbances and proposes a data-driven approach for predicting it by using machine learning techniques like Nonlinear Auto-regressive (NAR) Neural Networks (NN) and Long Short Term Memory (LSTM) networks from simulated and measured Phasor Measurement Unit (PMU) data. The proposed method uses a horizon-window that reconstructs the frequency input time-series data in order to predict the frequency features such as Nadir. Simulated scenarios are based on the gradual inertia reduction by including non-synchronous generation into the Nordic 32 test system, whereas the PMU collected data is taken from different locations in the Nordic Power System (NPS). Several horizon-windows are experimented in order to observe an adequate margin of prediction. Scenarios considering noisy signals are also evaluated in order to provide a robustness index of predictability. Results show the proper performance of the method and the adequate level of prediction based on the Root Mean Squared Error (RMSE) index.

show abstract

“…The LSTM network was developed to deal with the exploding and vanishing gradient problems encountered during the training of traditional RNNs, as well as to solve the long-term sequence dependence problem of RNNs. Deep learning architecture consist of LSTM layer (network) can address difficult sequence problems in deep learning and achieves state-of-the-art results in time series prediction [22]. Figure 1.…”

Section: Deep Learning With Long Short-term Memory (Lstm) Networkmentioning

confidence: 99%

Worldwide and Regional Forecasting of Coronavirus (Covid-19) Spread using a Deep Learning Model

Direkoğlu

Şah

2020

Preprint

View full text Add to dashboard Cite

In December 2019, Covid-19 epidemic was identified in Wuhan, China. Covid-19 may cause fatality especially among elderly, and people with chronic health problems. After human to human transmissions of the disease, it has rapidly spread throughout China, and then the outbreak has reached to neighboring countries in Asia. Now, the spread of the virus is accelerating in the world, and increasing number of new cases has been reported daily in Europe, Middle East, Africa and America regions. Recently, World Health Organization (WHO) also announced Covid-19 as a Pandemic. As of 3 April, worldwide around more than 1 million cases and around 60,000 fatalities are reported. Thus, forecasting regional and worldwide outbreak size of Covid-19 is important in order to take necessary actions regarding to preparedness plans and mitigation interventions. In this work, we design a deep learning model, which is an effective artificial intelligence method, to provide regional and worldwide forecasts. Particularly for worldwide, our approach predicts the cumulative number of cases, cumulative number of deaths and daily new cases. For Europe and Middle East regions, we predict the cumulative number of cases, and for Mainland China we predict daily new cases and the cumulative number of deaths. We predict the next 10 days based on the previously reported actual time series data of Covid-19. For worldwide forecasts, we use the data provided by Worldometers. For Europe and Middle East forecasts, we use the data provided by World Health Organization, and for China Mainland forecasts, the data is obtained from Chinese Centre for Disease Control and Prevention. This is the first time that a deep learning model has been employed for Covid-19 spread prediction, solely based on the known reported cases of Covid-19. The proposed deep learning architecture consists of Long Short Term Memory (LSTM) layer, dropout layer, and fully connected layers to predict regional and worldwide forecasts. We evaluate our approach with Root Mean Square Error (RMSE) metric. For forecasting, we use the network models that give the minimum RMSE on the last 3 days of actual data. Networks, which achieves the minimum RMSE on the last 3 days, are used to predict the next 10 days. Every day, the spread and situations are changing. Our approach can take into account these real-time changes; the deep learning model can be re-trained with new daily data and perform real-time forecasting. Results show that the proposed deep learning model is promising, it can predict possible scenarios regionally and globally for the spread of Covid-19.

show abstract

Recurrent Neural Networks for Time Series Forecasting: Current status and future directions

Cited by 736 publications

References 39 publications

Forecasting of COVID-19 cases using deep learning models: Is it reliable and practically significant?

Forecasting of COVID-19 cases using deep learning models: Is it reliable and practically significant?

Data-Driven Trajectory Prediction of Grid Power Frequency Based on Neural Models

Worldwide and Regional Forecasting of Coronavirus (Covid-19) Spread using a Deep Learning Model

Contact Info

Product

Resources

About