The paper presents the first results of the BayWay (Bayesian Roadway) project by providing an empirical comparison of some algorithms for intelligent data mining and Bayesian network learning on artificial road accident datasets. The theoretical motivations of the project are outlined, the problems involved in the construction of statistical models of road accidents dealt with, and a hybrid approach (MIDA) introduced. The machine learning algorithms and the artificial datasets used in the comparison are then presented. The methodology and the metrics utilised in the experiment are illustrated and the main results are described. The paper ends by drawing some general conclusions about the techniques to be used for mining road accident datasets.
Abstract