This paper addresses the application of Artifiical Neural Netwokrs (ANNs) in the emerging Stock Exchange of Greece in comparison with the developed Geman market. An collection of univariate and multivariate ANN's that implement two variations of the back-propagation algorithm are utilised to construct a trading system. The parametwrs, profitability and risk of the trading systems are determined by applying various methods including worst-case Drawdown and Probability of Ruin Anlaysis. The empirical evidence induicate that the performance of the ANN rtrading systems is sginificantly above random chance and that of other investment ttrategies and that this performance can vary according to shifts in volatility and major international events. The emerging Stock Market of Greece is found to present greater potentials for the succesful implementation of ANN's, when compared to the mature German stock market.
展开▼