| Abstract [eng] |
Considerable scientific efforts are aimed at the research of prediction models and methods. More complex models provide a better fitting to historical data, as usual. However, the recent publications on overfitting show that, under some unfavorable conditions, the correlation between the past and present results is negative. This work is to investigate both the influence of overfitting and the observed weak and, under some conditions, negative relation of portfolio profits as well as the prediction accuracy of asset prices applying the developed stock exchange model (Virtual Stock Exchange). The last problem is regarded as more important, since it is almost an open area so far, in the field of financial optimization. The new and most important conclusion of this work is that minimal prediction errors do not necessarily provide maximal portfolio profits. This conclusion and its underlying results have an impact on the research in the investment optimization. A particular attention should be paid to direct search of robust trading rules, less sensitive to unpredictable market changes. The expert systems that simulate the stock exchange are a useful tool for financial market research. To perform direct real-life experimentation in the real stock market is not practical, since the past financial time series cannot be repeated while working in real-time. All the research results have been obtained with the help of the expert system (Virtual Stock Exchange), developed using the Java programming tools. |
