Multilayer Feedforward Neural Network Model and Box-Jenkins Model for Seasonal Load Forecasting

eng Ansari Education and Research Society Journal of Ultra Scientist of Physical Sciences December 2008 20 3 767 722 jusps-B 1403 article Multilayer Feedforward Neural Network Model and Box-Jenkins Model for Seasonal Load Forecasting Norizan Mohamed (norizan@umt.edu.my) 1 Maizah Hura Ahmad 2 Zuhaimy Ismail 2 Khairil Anuar Arshad 2 Mathematics Department, Faculty of Science and Technology, University, Malaysia, Terengganu (UNT), 20130 Kula Terengganu, Terengganu, Malaysia Institution Department of Mathematics, Faculty of Science, Universiti Teknologi, Malaysia - 81310 UTm Skudai JOHAR (MALASIA) Artificial neural networks (ANNs) have been extensively studied and have been used as time series forecasting method. When neural network is compared to seasonal ARIMA (SARIMA) model, SARIMA model outperforms neural networks model when seasonality in a series exist. This paper aims to investigate the effectiveness of preprocessing data in neural networks model. In this study, the trend is absent and only seasonality exists, hence we only applied deseasonalization as preprocessing data. The forecasting performances among these three models, i.e., the SARIMA model, the neural network model with raw data and the neural network models with preprocessing data are compared. Comparing the performances using the root mean squared error (RMSE), the mean absolute error (MAE) and mean absolute percentage error (MAPE), we find that neural networks with preprocessing data are able to capture seasonality but SARIMA still outperforms both two neural network models.   https://ultraphysicalsciences.org/paper/1403/ Load Forecasting Deseasonalization Seasonal Autoregressive Integrated Moving Average Artificial Neural Networks Multilayer Feed-forward Neural Network