Comparison of iterative and direct approaches for multi-steps ahead time series forecasting using adaptive Hybrid-RBF neural network

Mazlina Mamat, Salina Abdul Samad

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

Abstract

Most available forecasters were designed in non-adaptive approach whereby the forecasters' parameters were updated during training phase. Slightly different, this paper introduces an adaptive forecaster built from the Hybrid Radial Basis Function neural network, in which its parameters were updated continuously in real time using new data. To achieve this, two learning algorithms: Adaptive Fuzzy C-Means Clustering and Exponential Weighted Recursive Least Square were used to train the Hybrid Radial Basis Function in adaptive mode. The multi-steps ahead forecasting were achieved by using two approaches: iterative and direct. The performance of each approach is measured by the Root Mean Square Error and R2 test of the actual and forecasted output on two time series data: Mackey-Glass and Data Series A from Santa-Fe Competition. Simulation results show that the adaptive forecaster is able to produce accurate forecasting output for several steps ahead depending on the complexity of data. Simulation results also reveal that the direct approach overcomes iterative approach in long distance forecasting.

Original languageEnglish
Title of host publicationIEEE Region 10 Annual International Conference, Proceedings/TENCON
Pages2332-2337
Number of pages6
DOIs
Publication statusPublished - 2010
Event2010 IEEE Region 10 Conference, TENCON 2010 - Fukuoka
Duration: 21 Nov 201024 Nov 2010

Other

Other2010 IEEE Region 10 Conference, TENCON 2010
CityFukuoka
Period21/11/1024/11/10

Fingerprint

Time series
Neural networks
Mean square error
Learning algorithms
Glass

Keywords

  • Adaptive learning
  • Artificial neural networks
  • Direct approach
  • Iterative approach
  • Time-series forecasting

ASJC Scopus subject areas

  • Electrical and Electronic Engineering
  • Computer Science Applications

Cite this

Comparison of iterative and direct approaches for multi-steps ahead time series forecasting using adaptive Hybrid-RBF neural network. / Mamat, Mazlina; Abdul Samad, Salina.

IEEE Region 10 Annual International Conference, Proceedings/TENCON. 2010. p. 2332-2337 5685968.

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

Mamat, M & Abdul Samad, S 2010, Comparison of iterative and direct approaches for multi-steps ahead time series forecasting using adaptive Hybrid-RBF neural network. in IEEE Region 10 Annual International Conference, Proceedings/TENCON., 5685968, pp. 2332-2337, 2010 IEEE Region 10 Conference, TENCON 2010, Fukuoka, 21/11/10. https://doi.org/10.1109/TENCON.2010.5685968
Mamat, Mazlina ; Abdul Samad, Salina. / Comparison of iterative and direct approaches for multi-steps ahead time series forecasting using adaptive Hybrid-RBF neural network. IEEE Region 10 Annual International Conference, Proceedings/TENCON. 2010. pp. 2332-2337
@inproceedings{6dfaa4cdc6154b7689d274fe862a4e07,
title = "Comparison of iterative and direct approaches for multi-steps ahead time series forecasting using adaptive Hybrid-RBF neural network",
abstract = "Most available forecasters were designed in non-adaptive approach whereby the forecasters' parameters were updated during training phase. Slightly different, this paper introduces an adaptive forecaster built from the Hybrid Radial Basis Function neural network, in which its parameters were updated continuously in real time using new data. To achieve this, two learning algorithms: Adaptive Fuzzy C-Means Clustering and Exponential Weighted Recursive Least Square were used to train the Hybrid Radial Basis Function in adaptive mode. The multi-steps ahead forecasting were achieved by using two approaches: iterative and direct. The performance of each approach is measured by the Root Mean Square Error and R2 test of the actual and forecasted output on two time series data: Mackey-Glass and Data Series A from Santa-Fe Competition. Simulation results show that the adaptive forecaster is able to produce accurate forecasting output for several steps ahead depending on the complexity of data. Simulation results also reveal that the direct approach overcomes iterative approach in long distance forecasting.",
keywords = "Adaptive learning, Artificial neural networks, Direct approach, Iterative approach, Time-series forecasting",
author = "Mazlina Mamat and {Abdul Samad}, Salina",
year = "2010",
doi = "10.1109/TENCON.2010.5685968",
language = "English",
isbn = "9781424468904",
pages = "2332--2337",
booktitle = "IEEE Region 10 Annual International Conference, Proceedings/TENCON",

}

TY - GEN

T1 - Comparison of iterative and direct approaches for multi-steps ahead time series forecasting using adaptive Hybrid-RBF neural network

AU - Mamat, Mazlina

AU - Abdul Samad, Salina

PY - 2010

Y1 - 2010

N2 - Most available forecasters were designed in non-adaptive approach whereby the forecasters' parameters were updated during training phase. Slightly different, this paper introduces an adaptive forecaster built from the Hybrid Radial Basis Function neural network, in which its parameters were updated continuously in real time using new data. To achieve this, two learning algorithms: Adaptive Fuzzy C-Means Clustering and Exponential Weighted Recursive Least Square were used to train the Hybrid Radial Basis Function in adaptive mode. The multi-steps ahead forecasting were achieved by using two approaches: iterative and direct. The performance of each approach is measured by the Root Mean Square Error and R2 test of the actual and forecasted output on two time series data: Mackey-Glass and Data Series A from Santa-Fe Competition. Simulation results show that the adaptive forecaster is able to produce accurate forecasting output for several steps ahead depending on the complexity of data. Simulation results also reveal that the direct approach overcomes iterative approach in long distance forecasting.

AB - Most available forecasters were designed in non-adaptive approach whereby the forecasters' parameters were updated during training phase. Slightly different, this paper introduces an adaptive forecaster built from the Hybrid Radial Basis Function neural network, in which its parameters were updated continuously in real time using new data. To achieve this, two learning algorithms: Adaptive Fuzzy C-Means Clustering and Exponential Weighted Recursive Least Square were used to train the Hybrid Radial Basis Function in adaptive mode. The multi-steps ahead forecasting were achieved by using two approaches: iterative and direct. The performance of each approach is measured by the Root Mean Square Error and R2 test of the actual and forecasted output on two time series data: Mackey-Glass and Data Series A from Santa-Fe Competition. Simulation results show that the adaptive forecaster is able to produce accurate forecasting output for several steps ahead depending on the complexity of data. Simulation results also reveal that the direct approach overcomes iterative approach in long distance forecasting.

KW - Adaptive learning

KW - Artificial neural networks

KW - Direct approach

KW - Iterative approach

KW - Time-series forecasting

UR - http://www.scopus.com/inward/record.url?scp=79951623961&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=79951623961&partnerID=8YFLogxK

U2 - 10.1109/TENCON.2010.5685968

DO - 10.1109/TENCON.2010.5685968

M3 - Conference contribution

SN - 9781424468904

SP - 2332

EP - 2337

BT - IEEE Region 10 Annual International Conference, Proceedings/TENCON

ER -