A Hybrid Model for Forecasting Stock Market Indexes in Nigeria

Select Volume / Issue:

• Volume 11, Issue 4, July 2023

Year:

2023

Type of Publication:

Article

Keywords:

Hybrid, Regression, Neural Network, Forecasting, Stock Price Indexes, Time Series, JEL Classification B22, C11, C15, C45

Authors:

Augustine Davou Pwasong; Fulus Daniel Fom; Bwirdimma Dugul Gotep; Wash Reng

Journal:

IJISM

Volume:

11

Number:

4

Pages:

67-84

Month:

July

ISSN:

2347-9051

Abstract:

In this study, the daily average stock price indexes for three banks in Nigeria was modeled and measured using three techniques. These techniques include the multiple linear regression (MLR) method, the cascade onward backtransmission neural network (CFBNN) method plus an amalgamation of a multiple linear regression method with cascade onward backtransmission neural network (MLR-CFBNN) method. The hybrid MLR-CFBNN was trained and tested on the increment series of the daily average stock price indexes of the three banks which are essentially univariate daily time sequence observations. An increment sequence of the original time sequence observations in this perspective refers to the difference series of the data. To determine the stationarity of the increment sequence, the Dickey-Fuller (DF) test was employed and the result revealed that the series is stationary. A comparison of the outcomes from the amalgamated MLR−CFBNN model and from the standalone cascade onward backtransmission neural network (CFBNN) technique as well as the standalone multiple linear regression (MLR) technique was conducted. Without loss of generality the developed hybrid MLR−CFBNN model displayed superior forecasting performance over standalone CFBNN and MLR techniques. This trend was achieved by the employment of the error of the mean in absolute terms commonly referred to as “mean absolute error” and the error root mean square (ERMS). The simulation was made possible by a MATLAB software compiler version 8.03. The developed hybrid model in this study is insightful and a prompt to the Securities and Exchange Commission to predict policies that will be in tandem with speculating the dynamics of stock businesses in Nigeria.

Full text: IJISM_986_FINAL.pdf [Bibtex]