An Application Of Neural Network On Developing The Trading Prediction And Analysis System

There are many neural networks trading systems to analysis and make decisions in finance market, using various investment strategies to maximize profits. The most recent researches focus on the prediction of probability distribution. A neural network model based on Gaussian Mixture Models has been devised by Dirk Husmeier in 1997 in order to predict the conditional probability densities. To demonstrate the merit of estimating probability density functions, this paper use this model to forecast the EUR/USD time series and then make an investment decision.In this paper, the GM model is applied to construct a neural network for forecasting the EUR/USD exchange rate, using the rate time series from 01/01/1999 to 12/31/2003 to train this model. Then I use this model to predicate the EUR/USD time series from 01/01/2004 to 03/15/2005. The results show that Gaussian Mixture Models perform very well when taking advantage of the possibilities combining with the threshold and models committee.