國立高雄大學圖資館 |

登入

回首頁

Applications of Deep Learning in Time Series Forecasting and Modeling Nonlinear Partial Differential Equations.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Applications of Deep Learning in Time Series Forecasting and Modeling Nonlinear Partial Differential Equations.
作者:	Roy, Debneil Saha.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, 2022
面頁冊數:	83 p.
附註:	Source: Dissertations Abstracts International, Volume: 84-02, Section: B.
附註:	Advisor: Li, Xiaolin.
Contained By:	Dissertations Abstracts International84-02B.
標題:	Applied mathematics.
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=29260180
ISBN:	9798841799917

Applications of Deep Learning in Time Series Forecasting and Modeling Nonlinear Partial Differential Equations.
Roy, Debneil Saha.

Applications of Deep Learning in Time Series Forecasting and Modeling Nonlinear Partial Differential Equations. - Ann Arbor : ProQuest Dissertations & Theses, 2022 - 83 p.

Source: Dissertations Abstracts International, Volume: 84-02, Section: B.

Thesis (Ph.D.)--State University of New York at Stony Brook, 2022.

This item must not be sold to any third party vendors.

Deep neural networks have received tremendous attention over the last many years, revolutionizing the fields of computer vision, speech recognition, natural language processing, etc. In this work, we investigate the application of deep learning methods for two important scientific problems, namely time series forecasting (electricity load forecasting and air temperature forecasting) and modeling nonlinear partial differential equations.In both electricity load forecasting and air temperature forecasting, we investigate the applicability of various deep learning models like multi-layer perceptron (MLP), Long Short Term Memory (LSTM) and a combination of Convolutional Neural Network (CNN) and LSTM (CNN+LSTM), for short to medium range forecasting. Even though electricity load forecasting generally occurs with system level data, forecasting at the household level is much more challenging, since electricity consumption behaviour can be much more variable than those at aggregate levels. With a significant increase in data generated at the household level due to the introduction of Advanced Metering Infrastructure (AMI) systems, deep learning methods are well suited for this problem. We compare the performance of the aforementioned deep neural networks with other time series based models and show that the former models outperform the latter. Furthermore, since deep neural networks take a long time to train, we propose a novel k-nearest neighbor based Echo State Network (a variant of recurrent neural network) for load forecasting and show that it is much easier to train and has a great deal of accuracy, on par with CNN+LSTM. For air temperature forecasting, we show that a combination of two neural networks has a better forecasting accuracy than single ones, since it is better able to extract important information from multiple related weather variables.We also investigate how deep neural networks can be used to solve nonlinear partial differential equations (PDEs) by incorporating physical principles in the model. The idea is to see whether by such learning, the model can forecast future states of PDEs.

ISBN: 9798841799917Subjects--Topical Terms:

377601
Applied mathematics.
Subjects--Index Terms:

Deep learning

Applications of Deep Learning in Time Series Forecasting and Modeling Nonlinear Partial Differential Equations.
LDR:03345nmm a2200373 4500 001 636144
005 20230501063910.5
006 m o d
007 cr#unu||||||||
008 230724s2022 ||||||||||||||||| ||eng d
020 $a 9798841799917
035 $a (MiAaPQ)AAI29260180
035 $a AAI29260180
040 $a MiAaPQ $c MiAaPQ
100 1 $a Roy, Debneil Saha. $3 942490
245 1 0 $a Applications of Deep Learning in Time Series Forecasting and Modeling Nonlinear Partial Differential Equations.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2022
300 $a 83 p.
500 $a Source: Dissertations Abstracts International, Volume: 84-02, Section: B.
500 $a Advisor: Li, Xiaolin.
502 $a Thesis (Ph.D.)--State University of New York at Stony Brook, 2022.
506 $a This item must not be sold to any third party vendors.
520 $a Deep neural networks have received tremendous attention over the last many years, revolutionizing the fields of computer vision, speech recognition, natural language processing, etc. In this work, we investigate the application of deep learning methods for two important scientific problems, namely time series forecasting (electricity load forecasting and air temperature forecasting) and modeling nonlinear partial differential equations.In both electricity load forecasting and air temperature forecasting, we investigate the applicability of various deep learning models like multi-layer perceptron (MLP), Long Short Term Memory (LSTM) and a combination of Convolutional Neural Network (CNN) and LSTM (CNN+LSTM), for short to medium range forecasting. Even though electricity load forecasting generally occurs with system level data, forecasting at the household level is much more challenging, since electricity consumption behaviour can be much more variable than those at aggregate levels. With a significant increase in data generated at the household level due to the introduction of Advanced Metering Infrastructure (AMI) systems, deep learning methods are well suited for this problem. We compare the performance of the aforementioned deep neural networks with other time series based models and show that the former models outperform the latter. Furthermore, since deep neural networks take a long time to train, we propose a novel k-nearest neighbor based Echo State Network (a variant of recurrent neural network) for load forecasting and show that it is much easier to train and has a great deal of accuracy, on par with CNN+LSTM. For air temperature forecasting, we show that a combination of two neural networks has a better forecasting accuracy than single ones, since it is better able to extract important information from multiple related weather variables.We also investigate how deep neural networks can be used to solve nonlinear partial differential equations (PDEs) by incorporating physical principles in the model. The idea is to see whether by such learning, the model can forecast future states of PDEs.
590 $a School code: 0771.
650 4 $a Applied mathematics. $3 377601
650 4 $a Information science. $3 190425
653 $a Deep learning
653 $a Forecasting
653 $a Partial differential equations
690 $a 0364
690 $a 0723
690 $a 0800
710 2 $a State University of New York at Stony Brook. $b Applied Mathematics and Statistics. $3 857349
773 0 $t Dissertations Abstracts International $g 84-02B.
790 $a 0771
791 $a Ph.D.
792 $a 2022
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=29260180