SET INDEX PREDICTION USING ARTIFICIAL NEURAL NETWORK

Abstract

In this study the ability of artificial neural network (ANN) in forecasting the daily SET stock exchange rate is investigated. 100 Non-linear Autoregressive Network with Exogenous Inputs (NARX) models along with two groups of five-prior days and ten-prior days have been assessed. The series of testing for 4 rounds of 4 different testing objectives, network architecture, training algorithm, hidden transfer function and output transfer function, are the methodology used in this study. The period of dataset is from January 27, 2005 till May 22, 2017 with totally 2920 time-step observations and the first 2828 observations are treated as training dataset and the next observations are treated as prediction dataset for testing the model prediction ability.

The results of study show that NARX can be used to forecast SET index accordingly and effectively along with 5 input parameters: daily net buy/sell of Thai institute investors, daily net buy/sell of foreign investors, daily Thai-US currency exchange rate (THB/USD), daily closing price of WTI price index (WTI) and daily closing price of US Dow Jones industrial stock index, and the model performance for prediction is 90.493% of the value of the coefficient of determination for regression plot.