Predicting trading signals of Sri Lankan stock market using Genetic Algorithms and Neural Networks

Abstract

This study predict the trading signals of Sri Lankan stock market using two sophisticated machine learning techniques called Genetic Algorithm (GA) and Neural Networks. These two techniques in combination predict the direction (going up or not) of the close price of tomorrow's (day t+1) 'All Share Price Index' (ASPI) of the Colombo Stock Exchange (CSE). The study period considered was from 1st November 2002 to 31st December 2008. The influential factors considered in this study represent the intermarket influence, political and environmental factors, economic stability and microeconomic factors: such as interest rate and exchange rate. A software called 'genetic model' was developed to find the optimum input variable combination that will affect the direction of tomorrow's ASPI value. Two identical neural network models called A and B were created for two different time periods, to predict the direction of ASPI of day (t+1). © Springer Science+Business Media B.V. 2010.