Predicting The Performance of Emerging Financial Markets Using NARX-NAR Neural Network

Abstract

One of the biggest challenges facing investors at the international level is measuring the performance of their investment operations within the structure of an international investment portfolio by predicting the future performance of its standard index.
In this research, we studied and analyzed the component prices of the MSCI Emerging International Index (Open-Max-Min-Close) as a time series based on testing several methods for modeling the closing price and comparing them with each other based on artificial neural networks for non-linear autoregressive NAR and NARX. In order to create a highly accurate neural model capable of responding quickly to price changes dynamically by predicting the future closing price of the general index, which represents the standard market portfolio, to support the process of making trading decisions in emerging markets.
The study concluded that it is possible to build a model to predict the future price of the general index for emerging financial markets based on several variables related to its component prices as a time series using the NARX neural network, with more accurate results than the model that was built based on a single variable, which is the closing price, as a time series using a network. Neural NAR, where the models were built based on the Matlab program, and the models’ effectiveness and generalizability were confirmed based on the mean square error (MSE) between the actual and expected values of the closing price, as the MSE value for the NARX network model was 0.00012653 and the MSE value for the NAR network model was 0.00015258.