Data Mining Technologies in Prediction of Stock Exchange

OData support
Petróczi Attila István
Department of Telecommunications and Media Informatics

The economic and the financial sector are using the possibilities offered by the information technologies and within it the data mining in an increasingly significant way. One of the hottest area in which these methods are applied is the stock exchange environment. In this segment these methods are used to predict the future movement and risks of the securities as precisely as possible.

In this paper I will analyze and test data mining methods that can be applied in the prediction of the securities. I will make an effort to predict these time series and to develop portfolio managerial procedures that would be able to profit from the forecasts in a stock exchange environment.

These tasks usually can not be solved effectively with conventional, sequential programming algorithms, so it worth to study the methods of the peripheral areas of the information technology, like the artificial intelligence researches. In my paper I am using up two of these methods, the artificial neural networks and the genetic algorithms.

The neural networks have proven their usability in many areas, and for some problems such as pattern recognition they seem to be the best solution. In data mining the feed-forward and many kind of recurrent versions of the networks are used for time series prediction.

I use the genetic algorithms for the unsupervised training of the neural networks. This method is especially useful in the developing procedure of the portfolio managing methods, since the perfect trading processes for the different quality forecasts are not known, so an unsupervised training process must be used.

In the paper I first review the theoretical background of the neural networks and the genetic algorithms, and then examine the possible models for the prediction and portfolio managing tasks. Finally I test the developed methods and evaluate the results.


Please sign in to download the files of this thesis.