Comparison of neural network and regression techniques for nonlinear prediction problems

The aim of this study is to compare the predictive performance of feed forward neural network with some of the regression models that are capable of handling certain nonlinear prediction problems. Four real life examples are considered in this study where the response variable belongs to the exponential family of distributions and are modeled using generalized linear models. Results point out the merit of using appropriate regression models when the functional relationship between the variables is known apriori.