International Journal on Advanced Science, Engineering and Information Technology, Vol. 7 (2017) No. 3, pages: 1098-1103, DOI:10.18517/ijaseit.7.3.1363

Multi-Layer Perceptron (MLP)-Based Nonlinear Auto-Regressive with Exogenous Inputs (NARX) Stock Forecasting Model

I. M. Yassin, M. F. Abdul Khalid, S. H. Herman, I. Pasya, N. Ab Wahab, Z. Awang

Abstract

The prediction of stocks in the stock market is important in investment as it would help the investor to time buy and sell transactions to maximize profits. In this paper, a Multi-Layer Perceptron (MLP)-based Nonlinear Auto-Regressive with Exogenous Inputs (NARX) model was used to predict the prices of the Apple Inc. weekly stock prices over a time horizon of 1995 to 2013. The NARX model belongs is a system identification model that constructs a mathematical model from the dynamic input/output readings of the system, and predicts the future behaviour of the system based on the constructed mathematical model. The One Step Ahead (OSA) and correlation tests were used to test validate the model. Results demonstrate the predictive ability of the model while producing Gaussian residuals (indicating the validity of the model).

Keywords:

system identification; stock market forecasting; nonlinear auto-regressive with exogenous inputs (NARX), multi-layer perceptron (MLP)

Viewed: 446 times (since Sept 4, 2017)

cite this paper     download