As accidents tend to be multicausal, the interpretation of accident data can be a fairly complex task. Hence it is worth experimenting with innovative procedures in order to extrapolate patterns within such data. Accordingly, records of motorway accidents in northern Italy, stored on statistical cards, were processed by means of a neural network. The clustering ability of the latter allowed for an interpretive assessment of each input variable in terms of its influence on the number of accidents occurring. (A)
Abstract