Application of Genetic Algorithms to the Optimisation of Neural Network Configuration for Stock Market Forecasting

摘要

Neural networks are recognised as an effective tool for predicting stock prices (Shin & Han, 2000), but little is known about which configurations are best and for which indices. The present study uses genetic algorithms to find a near optimal learning rate, momentum, tolerance and network architecture for 47 indices listed on the Australian Stock Exchange (ASX). Some relationships were determined between stock index and neural network attributes, and important observations were made for the further development of a methodology for determining optimal neural network configurations.