Journal of Management Information System & E-commerce

A Performance Comparison between Multiple Linear Regression and Artificial Neural Networks in Predicting the Number of New Business Registered: A Case of the Democratic Republic of the Congo.
Mulomba Mukendi Christian

Abstract
The new business registered (numbers) is the number of new limited liability corporations registered in the calendar year. Established in 2006 by the World Bank Group, this indicator can be used to determine factors impacting private sector growth and to measure the level of entrepreneurship in a country. The aim of this research is to use information and communication technology to predict this indicator in comparing the performances of multiple linear regression and artificial neural networks for the case of the Democratic Republic of Congo. After the finding of some factors common to many countries from the literature and collecting data over 27 countries for each factors (variables) from various sources (secondary data), a model building and prediction on the targeted country took place using Microsoft Excel 2013 for Multiple Linear Regression and the Neural Tool of Palisade suite 7.5 for Artificial Neural Network. Results shows that neural networks performs better prediction than multiple linear regression and thus can provide accurate prediction if used with significant variables. A decision support system using this model can be implemented and upgrade in order to meet this need for the country.

Full Text: PDF     DOI: 10.15640/jmise.v5n2a1