Simulation of Regression Analysis by an Automated System utilizing Artificial Neural Networks

Abstract

Artificial Neural Networks have been gaining popularity as statistical tools since it resolves some disadvantages of conventional regression analysis techniques. This paper describes the implementation issues on designing dynamically changing artificial neural networks which are to be applied for the situations where the Regression Analysis is to be used. Furthermore, in order to resolve some of the problems of existing statistical packages like MINITAB, R and SAS, a computer based analysis system is proposed in order to simulate the complete process of building up a regression model and to make future predictions. When implementing the automated system, we used JAVA which supports Object Oriented Programming and MATLAB for easy calculation of mathematical functions. Finally we present a comparative study on the results obtained by the proposed system and the conventional statistical methods. This system provides better output in identifying relationships between independent and dependent variables compared to conventional regression techniques.