An application of discriminant analysis and artificial neural networks to classification problems

In this paper we present three algorithms to classify the elements of a population into two groups. The first algorithm is the classical multivariate statistical technique – discriminant analysis using Bayesian decision theory for the classification problem. The other two algorithms are two types of artificial neural networks: a feedforward three-layer network with a back-propagation learning algorithm and the probabilistic neural network. After discussing some theoretical attributes of these models, emphasis is placed on empirical results.