The predictive accuracy of feed forward neural networks and multiple regression in the case of heteroscedastic data

During the last few years, several comparative studies for regression analysis and neural networks have been published. Our paper contributes to this stream of research by comparing the performance of feed forward neural network and multiple regression when heteroscedasticity is present in the data. Datasets are simulated that vary systematically on various dimensions like sample size, noise levels and number of independent variables to assess the consequences of deviations from underlying assumptions of homoscedasticity on the comparative performance of regression analysis and neural networks. Comparative analysis is carried out using appropriate experimental design and the results are presented.