ABSTRACT

This study aims to examine performance of artificial neural networks (ANN) in forecasting stock market index movement. The forecasting is based on two samples of Istanbul Stock Exchange (ISE) data and each consisting of 150 observations. Forecasting performance is assessed in one, five and ten day periods. Results show that ANN gives high percentage of correctly forecasted signs. This performance is particularly evident in five days period, while conventional approach mostly uses next day forecasting. These results imply that for more efficient forecasting with ANN use of different time periods is important.