Warianty tytułu
Analysis of Influence of Initial Learning Parameters on the Neural Networks Effectiveness in Financial Forecasting
Języki publikacji
Abstrakty
W pracy zaprezentowano wyniki badań wpływu dobieranych losowo parametrów początkowych procesu uczenia sieci neuronowej na jej działanie w zagadnieniach prognozowania i wspomagania decyzji inwestycyjnych dla finansowych szeregów czasowych. W analizie zastosowano sieci neuronowe Bayesa do przewidywania kierunku zmian indeksu giełdowego WIG 20.
The article presents research results concerning the influence of randomly selected initial parameters of neural network learning process on the network effectiveness in the issues of financial time series forecasting and investment decision making. The Bayesian neural networks, constructed for the prediction of direction of stock index WIG 21 movement, have been applied in the research. The efficiency of investment strategies generated by the network has been evaluated and the analysis of relations between selected indicators of network performance, calculated for 100 different sets of random initial parameters in learning and test periods, has been carried out. The practical conclusion have been formulated and the future research concept has been proposed. (original abstract)
Rocznik
Numer
Strony
49-66
Opis fizyczny
Twórcy
autor
Bibliografia
- Azoff E.M. [1994], Neural Network Time Series Forecasting of Financial Markets, Wiley, New York.
- Haykin S. [1994], Neural Networks. A Comprehensive Foundation, Macmillan College Publishing Company, New York.
- Inteligentne systemy w zarządzaniu - teoria i praktyka [2000], J.S. Zieliński (red.), Wydawnictwo Naukowe PWN, Warszawa.
- Lula P., [1999], Jednokierunkowe sieci neuronowe w modelowaniu zjawisk ekonomicznych, Wydawnictwo AE w Krakowie, Kraków.
- MacKay D.J.C. [1992a], Bayesian Interpolation, Neural Computation, 4.
- MacKay D.J.C. [1992b], A Practical Bayesian Frameworkfor Backpropagation Networks, Neural Computation 4.
- Morajda J., [1999], Applications of Neural Networks in the Financial Markets - Selected Aspects, Proc. of the 4th Conference Neural Networks and Their Applications, Zakopane.
- Morajda J., [2000], Neural Networks as Predictive Models in Financial Futures Trading, Proc. of the 5th Conference Neural Networks and Soft Computing, Zakopane. Neural networks in the capital markets [1995], A.P. Refenes (ed.), Chichester, Wiley.
- Osowski S. [1996], Sieci neuronowe w ujęciu algorytmicznym, WNT, Warszawa.
- Tadeusiewicz R. [1993], Sieci neuronowe, Akademicka Oficyna Wydawnicza RM, Warszawa.
Typ dokumentu
Bibliografia
Identyfikatory
Identyfikator YADDA
bwmeta1.element.ekon-element-000042037626