A Principled Approach to Building and Evaluating Neural Network Models for E-Business Applications

Every commercial transaction generates large amounts of data on consumers for use by organizations. Data from ebusiness is typified by its complexity, quantity, and noisiness. Neural networks are ideally suited for these problem characteristics. Furthermore, the fact that neural networks can estimate the posterior probabilities associated with the group membership of objects of interest, makes them a powerful tool of great potential for e-business applications. As with all classification approaches, though, the neural network’s utility is based upon its generalization performance on new data. In this paper, we propose a principled approach to building and evaluation neural network models for e-business applications. First, the usefulness of neural networks for e-commerce applications and Bayesian classification is discussed. Next, the theory concerning model accuracy and generalization is presented. Then the principled approach is described including illustrative examples