VESSEL ACCIDENT MODELING: A COMPARISON OF NEURAL NETWORKS, DISCRIMINANT ANALYSIS AND LOGISTIC REGRESSION

Accidents are, by definition, unplanned and unforeseen occurrences. Their causes are often circumstantial, with chance serving as a catalyst. But accidents are not necessarily improbable occurrences. Indeed, circumstances may provide any degree of likelihood that they will occur. This study investigates the circumstances surrounding navigational accidents and evinces some significant conclusions about the likelihood associated with such events. As applied in this research, neural network analysis assumes that human learning can be emulated by a network of massively interconnected but very simple processing units. In this study, a neural network is employed to predict the type of vessel accident that might occur along a major waterway - the lower Mississippi River from above Baton Rouge to the Gulf of Mexico.