4 Hours NNARX flood prediction model using 'traingd' and 'trainoss' training function: A comparative study

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Research on flood disasters are popular area of researches and more and more advanced tools are uses to obtain reliable predicted results. This due to flood causes harm to people and their property specifically. Thus, many hydrological models are propose and developed by researchers around the world to forecast flood disaster ahead of time for evacuation purposes. However, it is very difficult to develop flood model because all physical parameters that represent the flood behaviour must be included in the modelling. Parameters such as the depth of river basin, river flow rate and sediment factors data are very difficult to get. Hence, this paper proposes flood prediction model using black-box model. It is call black-box model because the model is develop using only input and output data without the needs to know the physical parameters that contribute to flood disaster. Neural Network Autoregressive with Exogenous Inputs (NNARX) is one type of black-box model that is widely applied to solve the nonlinear problems. In order to develop the model, flood data were divide into 3 samples that are modelling sample, validation sample and testing sample. All samples were provide from related government agency and it is real-time data. The model were develop using two different training functions in order to obtain the optimal ones. The prediction time is set at 4 hours ahead of time. Simulation works and model performances was done using MATLAB Neural Network Toolbox. Findings show that the propose NNARX model with 4 hours prediction time that trained using 'traingd' had produced better result.

LanguageEnglish
Title of host publicationProceedings - 2018 IEEE 14th International Colloquium on Signal Processing and its Application, CSPA 2018
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages77-81
Number of pages5
ISBN (Electronic)9781538603895
DOIs
Publication statusPublished - May 29 2018
Event14th IEEE International Colloquium on Signal Processing and its Application, CSPA 2018 - Penang, Malaysia
Duration: Mar 9 2018Mar 10 2018

Other

Other14th IEEE International Colloquium on Signal Processing and its Application, CSPA 2018
CountryMalaysia
CityPenang
Period3/9/183/10/18

Fingerprint

Disasters
Rivers
Neural networks
Catchments
MATLAB
Sediments
Flow rate
Testing

All Science Journal Classification (ASJC) codes

  • Computer Science Applications
  • Computer Vision and Pattern Recognition
  • Signal Processing

Cite this

Ruslan, F. A., Samad, A. M., & Adnan, R. (2018). 4 Hours NNARX flood prediction model using 'traingd' and 'trainoss' training function: A comparative study. In Proceedings - 2018 IEEE 14th International Colloquium on Signal Processing and its Application, CSPA 2018 (pp. 77-81). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/CSPA.2018.8368689

4 Hours NNARX flood prediction model using 'traingd' and 'trainoss' training function : A comparative study. / Ruslan, Fazlina Ahmat; Samad, Abd Manan; Adnan, Ramli.

Proceedings - 2018 IEEE 14th International Colloquium on Signal Processing and its Application, CSPA 2018. Institute of Electrical and Electronics Engineers Inc., 2018. p. 77-81.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Ruslan, FA, Samad, AM & Adnan, R 2018, 4 Hours NNARX flood prediction model using 'traingd' and 'trainoss' training function: A comparative study. in Proceedings - 2018 IEEE 14th International Colloquium on Signal Processing and its Application, CSPA 2018. Institute of Electrical and Electronics Engineers Inc., pp. 77-81, 14th IEEE International Colloquium on Signal Processing and its Application, CSPA 2018, Penang, Malaysia, 3/9/18. https://doi.org/10.1109/CSPA.2018.8368689
Ruslan FA, Samad AM, Adnan R. 4 Hours NNARX flood prediction model using 'traingd' and 'trainoss' training function: A comparative study. In Proceedings - 2018 IEEE 14th International Colloquium on Signal Processing and its Application, CSPA 2018. Institute of Electrical and Electronics Engineers Inc. 2018. p. 77-81 https://doi.org/10.1109/CSPA.2018.8368689
Ruslan, Fazlina Ahmat ; Samad, Abd Manan ; Adnan, Ramli. / 4 Hours NNARX flood prediction model using 'traingd' and 'trainoss' training function : A comparative study. Proceedings - 2018 IEEE 14th International Colloquium on Signal Processing and its Application, CSPA 2018. Institute of Electrical and Electronics Engineers Inc., 2018. pp. 77-81
@inproceedings{1515ca4f9f1449358d5cd83f8f557b12,
title = "4 Hours NNARX flood prediction model using 'traingd' and 'trainoss' training function: A comparative study",
abstract = "Research on flood disasters are popular area of researches and more and more advanced tools are uses to obtain reliable predicted results. This due to flood causes harm to people and their property specifically. Thus, many hydrological models are propose and developed by researchers around the world to forecast flood disaster ahead of time for evacuation purposes. However, it is very difficult to develop flood model because all physical parameters that represent the flood behaviour must be included in the modelling. Parameters such as the depth of river basin, river flow rate and sediment factors data are very difficult to get. Hence, this paper proposes flood prediction model using black-box model. It is call black-box model because the model is develop using only input and output data without the needs to know the physical parameters that contribute to flood disaster. Neural Network Autoregressive with Exogenous Inputs (NNARX) is one type of black-box model that is widely applied to solve the nonlinear problems. In order to develop the model, flood data were divide into 3 samples that are modelling sample, validation sample and testing sample. All samples were provide from related government agency and it is real-time data. The model were develop using two different training functions in order to obtain the optimal ones. The prediction time is set at 4 hours ahead of time. Simulation works and model performances was done using MATLAB Neural Network Toolbox. Findings show that the propose NNARX model with 4 hours prediction time that trained using 'traingd' had produced better result.",
author = "Ruslan, {Fazlina Ahmat} and Samad, {Abd Manan} and Ramli Adnan",
year = "2018",
month = "5",
day = "29",
doi = "10.1109/CSPA.2018.8368689",
language = "English",
pages = "77--81",
booktitle = "Proceedings - 2018 IEEE 14th International Colloquium on Signal Processing and its Application, CSPA 2018",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - GEN

T1 - 4 Hours NNARX flood prediction model using 'traingd' and 'trainoss' training function

T2 - A comparative study

AU - Ruslan, Fazlina Ahmat

AU - Samad, Abd Manan

AU - Adnan, Ramli

PY - 2018/5/29

Y1 - 2018/5/29

N2 - Research on flood disasters are popular area of researches and more and more advanced tools are uses to obtain reliable predicted results. This due to flood causes harm to people and their property specifically. Thus, many hydrological models are propose and developed by researchers around the world to forecast flood disaster ahead of time for evacuation purposes. However, it is very difficult to develop flood model because all physical parameters that represent the flood behaviour must be included in the modelling. Parameters such as the depth of river basin, river flow rate and sediment factors data are very difficult to get. Hence, this paper proposes flood prediction model using black-box model. It is call black-box model because the model is develop using only input and output data without the needs to know the physical parameters that contribute to flood disaster. Neural Network Autoregressive with Exogenous Inputs (NNARX) is one type of black-box model that is widely applied to solve the nonlinear problems. In order to develop the model, flood data were divide into 3 samples that are modelling sample, validation sample and testing sample. All samples were provide from related government agency and it is real-time data. The model were develop using two different training functions in order to obtain the optimal ones. The prediction time is set at 4 hours ahead of time. Simulation works and model performances was done using MATLAB Neural Network Toolbox. Findings show that the propose NNARX model with 4 hours prediction time that trained using 'traingd' had produced better result.

AB - Research on flood disasters are popular area of researches and more and more advanced tools are uses to obtain reliable predicted results. This due to flood causes harm to people and their property specifically. Thus, many hydrological models are propose and developed by researchers around the world to forecast flood disaster ahead of time for evacuation purposes. However, it is very difficult to develop flood model because all physical parameters that represent the flood behaviour must be included in the modelling. Parameters such as the depth of river basin, river flow rate and sediment factors data are very difficult to get. Hence, this paper proposes flood prediction model using black-box model. It is call black-box model because the model is develop using only input and output data without the needs to know the physical parameters that contribute to flood disaster. Neural Network Autoregressive with Exogenous Inputs (NNARX) is one type of black-box model that is widely applied to solve the nonlinear problems. In order to develop the model, flood data were divide into 3 samples that are modelling sample, validation sample and testing sample. All samples were provide from related government agency and it is real-time data. The model were develop using two different training functions in order to obtain the optimal ones. The prediction time is set at 4 hours ahead of time. Simulation works and model performances was done using MATLAB Neural Network Toolbox. Findings show that the propose NNARX model with 4 hours prediction time that trained using 'traingd' had produced better result.

UR - http://www.scopus.com/inward/record.url?scp=85048798229&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85048798229&partnerID=8YFLogxK

U2 - 10.1109/CSPA.2018.8368689

DO - 10.1109/CSPA.2018.8368689

M3 - Conference contribution

SP - 77

EP - 81

BT - Proceedings - 2018 IEEE 14th International Colloquium on Signal Processing and its Application, CSPA 2018

PB - Institute of Electrical and Electronics Engineers Inc.

ER -