Volume 18, Number 1, June 2023
| Comprehensive Analysis of Neural Network with GARCH for Predicting the TAIEX Futures |
Abstract
From the viewpoint of investors, the higher the accuracy is for share price predictions, the lower their investment risks and hence the greater the returns will be. For companies, share prices reflect their future value and current operational status and showcase their strengths and weaknesses. This paper thus proposes a combination of the cost-of-carry pricing model with the back propagation neural network (BPNN) for predictions made with time-series data of futures and based on the variables of the cost-of-carry pricing model. We perform a comparison of the predictions and the results derived with the cost-of-carry pricing model in order to identify a reasonable and accurate forecast model. The empirical findings suggest that the BPNN price model shows superior price discovery capability than the traditional cost-of-carry pricing model. In other words, investors can reasonably and accurately predict prices by using BPNN to enhance their probability of achieving excess returns.
Keywords: Artificial Neural Network, Back Propagation neural network, Prineipal Component Analysis, Volatility
JEL Classification: G12, G14, G17