Neural network system for forecasting method selection

This paper first examines the general technical issues (decisions) involved in designing neural network applications, then a backpropagation-based forecasting prototype is used to demonstrate how these decisions be determined in practice.

Abstract

Choosing an appropriate forecasting method is a crucial decision for most organizations, as the company's success is highly dependent on the accurate prediction of future. The decision, however, is not easy because many forecasting methods are available and the selection often requires extensive statistical knowledge, and personal judgment. In this paper, we illustrate how can a neural network approach be used to ease this task. We first examine the general technical issues (decisions) involved in designing neural network applications. A backpropagation-based forecasting prototype is then used to demonstrate how these decisions be determined in practice.