34 – Neural Networks for Economic Forecasting Problems

Publisher Summary Since neural networks can learn nonlinear relationships between inputs and desired outputs, predicting the future behavior of real-world time series using neural networks has been extensively examined. Numerous attempts have been made to apply neural networks to financial market prediction, electricity-load forecasting, and other areas such as flour price prediction. The aim of this chapter is to introduce various neural network approaches for economic forecasting problems such as univariate time-series forecasting, multivariate prediction, hybrid systems, and recurrent neural networks. This chapter illustrates multivariate stock market prediction by examining the stock market prediction of Germany, UK, and Japan. Multivariate prediction results are better than those from univariate time-series forecasting. Prediction results can be improved with the use of knowledge. Learning techniques such as weight elimination and selective presentation can improve the network performance. An important area for further study is hybrid systems and the ways to integrate statistical analysis and neural networks.