Financial Forecasting Using Neural Networks
Dagmar Blatná, Jiří Trešl · 2011
The possibility of application of neural networks for the prediction of both stock and exchange rate FCLUIIIS Was investigated. First, the capability of neural networks to reveal specific underlying process was studied using different simulated time series. Second, actual weekly returns from Czech financial markets were analyzed and predicted. Particularly, the problems connected with capturing of outliers and structural breaks were discussed. The predictive power of neural networks was investigated both as a function of network architecture and the length of training set.