Neural networks of combination of forecasts for data with long memory pattern

In the current paper, we experiment with three neural network models for forecasting to better understand the performance of neural networks for the case when the data exhibits a long memory pattern. The first model is a simple one consisting of the time series values. The second model uses the inputs as the first model with one additional input representing the combination (average) of five univariate time series forecasting models. The third model consists of all these individual time series forecasts, in which the neural network is let to combine them in its own way. The network characteristics such as the training parameters, the number of hidden layers, and the testing and training percentages are simulated. The univariate forecasting models used by the neural network include Box-Jenkins and exponential smoothing. The exponential smoothing models include the simple exponential smoothing, linear multiplicative, linear additive, decaying linear multiplicative, and decaying linear additive types. Neural network simulations were performed using the feedforward, back propagation procedure. Electricity peak-load data for the Emirate of Abu-Dhabi is used for the experiments