Neural Networks to Predict Financial Time Series in a Minority Game Context

In this paper we consider financial time series from U.S. Fixed Income Market, S&P500, Exchange Market and Oil Market. It is well known that financial time series reveal some anomalies as regards the Efficient Market Hypotesis and some scaling behavior is evident such as fat tails and clustered volatility. This suggests to consider financial time serie as "pseudo"-random time series. For this kind of time series the power of prediction of neural networks has been shown to be appreciable. We first consider the financial time serie from the Minority Game point of view and than we apply a neural network with learning algorithm in order to analyze its prediction power. We show that Fixed Income Market presents many differences from other markets in terms of predictability as a measure of market efficiency.