Fitting conventional neural network time series models on sand price indices dataset

• Main Authors: Ghani, N.A.M (Author), Hashim, H. (Author), Kamaruddin, S.B.A (Author), Musirin, I. (Author)
• Format: Article
• Language: English
• Published: Science Publishing Corporation Inc 2018
• Subjects: Conventional Nonlinear Autoregressive (NAR); Conventional Nonlinear Autoregressive Moving Average (NARMA); Neural Network; Sand Price Indices
• Online Access: View Fulltext in Publisher; View in Scopus
• ISBN: 2227524X (ISSN)
• DOI: 10.14419/ijet.v7i3.15.17396

This result-based paper discusses on the best aftereffects of both fitted BPNN-NAR and BPNN-NARMA on MCCI Sand dataset regarding distinctive error measures. This exploration examine the outcomes as far as the execution of the fitted forecasting models by every arrangement of input lags and error lags utilized, the execution of the fitted anticipating models by various hidden nodes utilized, the execution of the fitted estimating models when joining both inputs and hidden nodes, the consistency of error measures utilized for the fitted determining models, and in addition the general best fitted estimating models for Malaysian sand price indices dataset. In this examination, Malaysian sand price indices monthly data from January 1980 to December 2013 were adapted. The examination of BPNN-NAR on Malaysian sand data shows that insufficient or inadequate combination of input and error lags lead to greater RMSE. Correspondingly, the number of input lags for BPNN-NAR, as well as the number of input and error lags for BPNN-NARMA really have direct effect to the models' performances. The higher or the lesser hidden nodes to the input lags, the higher the network's RMSE. On the other hand, the higher the input lags lead to the higher network's RMSE. © 2018 Authors.