Evaluation of neural network forecasting methods : a comparative study with application

Abstract EN

A time series is a time-ordered sequence of observations of a variable.

A problem which is of great importance is the prediction of future values of a time series and the necessity of a good method of forecasting. This study aims to forecast the future values of the Egyptian exchange industries using a recursive linear regression modeling approach, and by using a neural network approach as an ideal choice for flexible nonlinear functional approximation.

The nonlinear neural network model is found to have better results in sample fit and forecasts compared to its linear counterpart.

A multilayer feedforward network is constructed and trained by back-propagation algorithm. This paper also implements different methods for using neural networks in forecasting problems, these methods were applied to preprocess the output of the neural network for single step-prediction and multi-step prediction, to forecast the future values of exchange industries of Egypt directly, and the difference between the current period and the next period.