Simulating the process of multiattribute choice with neural networks

In most multiattribute utility models, it is assumed that the analytical forms of the partial and global utility functions are known, and the problem is then one of estimating the parameters of the model. It is shown that a class of neural networks can be used to solve this multiattributable choice problem. The process of multiattribute choice involving a set of alternatives is discussed. An equivalence between the multiattribute utility model, as presented, and an a posteriori probability model, which can be computed by a neural network, is established. This result is applied to a small illustrative example for training a neural network