Title :
Energy function construction and implementation for stock exchange prediction NNs
Author :
Cristea, Alexandra I. ; Okamoto, Toshio
Author_Institution :
Graduated Sch. of Inf. Syst., Univ. of Electro-Commun., Tokyo, Japan
Abstract :
Neural networks (NN), with their parallel processing power, can be used as a tool to forecast stock exchange events (SEE), as a sub-domain of time-series (TS) forecasting. For the final product of SEE forecasts, other external economical factors have to be taken also into consideration and to be combined with the pure TS forecast. In this paper we present the energy function construction and implementation for SEE prediction. We focus on the mathematical deductions of the energy function and on the error minimization procedures. We present also some comparative results of our method, based on Lyapunov (also called infinite) norm, compared to the classical backpropagation method (BP), and to the random walk generator. We discuss some further optimisation of the system
Keywords :
Lyapunov methods; forecasting theory; neural nets; stock markets; time series; BP; Lyapunov norm; SEE; backpropagation; energy function; energy function construction; error minimization; external economical factors; infinite norm; neural networks; parallel processing; random walk generator; stock exchange events; stock exchange prediction; time-series forecasting; Artificial intelligence; Backpropagation; Chaos; Economic forecasting; Electronic mail; Information systems; Neural networks; Parallel processing; Power generation economics; Stock markets;
Conference_Titel :
Knowledge-Based Intelligent Electronic Systems, 1998. Proceedings KES '98. 1998 Second International Conference on
Conference_Location :
Adelaide, SA
Print_ISBN :
0-7803-4316-6
DOI :
10.1109/KES.1998.726001